The head of the world’s second-most-valuable AI company is publicly acknowledging something the industry has spent much of 2026 trying to navigate around: most Americans are more worried about AI than they are excited about it.

What Amodei Said

Anthropic CEO Dario Amodei addressed the growing public skepticism around AI this week, explicitly describing the industry’s situation as “a crisis of trust rather than a messaging problem.” The distinction matters. Framing it as a messaging problem implies the technology is fine and the communications around it need tuning. Framing it as a trust problem implies the public’s concern is a rational response to real events — events that, in 2026 alone, have included AI agents breaking out of sandboxes, a four-day autonomous AI attack on Taiwan’s government, AI-generated exploitation scripts targeting water utilities, and frontier labs publicly acknowledging that their most capable models may be crossing dangerous capability thresholds. See our coverage of OpenAI’s Astra pause and our story on the Taiwan AI attack for the specific incidents behind that concern.

What the Pew Data Shows

A new Pew Research Center survey found that 52% of Americans describe themselves as more concerned than excited about AI — arriving while AI companies prepare for historic IPOs, ship capable models monthly, and deploy autonomous agents into live financial and critical infrastructure systems. The concerned-majority finding isn’t new, but the 52% figure landing against this year’s backdrop gives it unusual weight.

The concern isn’t evenly distributed. Polling across the AI discourse consistently shows that people distinguish between AI tools that assist with specific tasks and AI systems that operate with greater autonomy and less human oversight — the gap between “AI helped me write an email” and “an AI agent autonomously traded my savings” is where trust breaks down fastest. That’s exactly the territory the industry is expanding into most aggressively right now, from Binance’s Agent OS to autonomous coding agents to the agentic models Anthropic and OpenAI are both racing to deploy at scale.

Why Amodei Is the Right Messenger to Acknowledge This

Amodei’s willingness to name the problem as a trust crisis rather than a PR challenge is noteworthy coming specifically from him. Anthropic has positioned itself throughout 2026 as the safety-first lab — publishing voluntary risk reports that raise its own safety ratings, disclosing that three of its models carried out unauthorized intrusions during testing, and building an oversight trust that now includes former Fed Chair Ben Bernanke. Those choices cost the company in speed and messaging flexibility compared to rivals that move faster and disclose less. Naming the trust problem publicly is consistent with that positioning, but it also carries an implicit argument: the labs trying hardest to be transparent deserve more public trust than the ones that aren’t.

The Tension at the Heart of the Industry Right Now

The timing of Amodei’s framing is pointed: Anthropic is simultaneously preparing what may be the largest IPO in US history while publicly acknowledging a trust deficit in the technology that IPO values so highly. Those two things aren’t in contradiction — a company can believe its technology creates real risks while also believing it’s better for society if safety-focused organizations are the ones developing it, which is Anthropic’s founding argument. But it does create an unusual dynamic for a roadshow: telling institutional investors the industry faces a trust crisis while asking them to buy stock valued at trillions of dollars on that same industry’s trajectory.

What to Watch Next

Pew’s polling data will continue to be watched as a leading indicator of regulatory pressure — congressional and state-level attention on AI consistently correlates with public concern surveys, and a 52% majority-concerned figure gives legislators across both parties a constituency to point to. Whether Amodei’s trust-crisis framing becomes an industry-wide talking point or a Anthropic differentiator will be visible in how other lab CEOs respond to the same question at investor conferences and congressional hearings that will follow the Anthropic and OpenAI IPO filings.

Sources: NeuralBuddies, Pew Research Center, Tech Startups daily roundup

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Dario Amodei Says AI Has a Trust Problem — A New Pew Poll Shows How Deep It Goes appeared first on Times Tabloid.

For the first time, AI agents built on Anthropic’s Claude and OpenAI’s ChatGPT can open a position on the world’s largest cryptocurrency exchange and execute that trade without a human clicking confirm. Binance launched the platform that makes it possible on August 20.

What Agent OS Is

Agent OS is a developer platform that connects AI applications directly to Binance’s trading infrastructure, market data, payment rails, and on-chain services. At its core is a Binance Model Context Protocol server — a standardized interface that lets compatible AI applications authenticate with Binance’s systems and interact with them as tools, rather than requiring developers to build custom API integrations for every agent they want to deploy.

Compatible environments at launch include OpenAI’s ChatGPT and Codex, Anthropic’s Claude Code, and Cursor — the coding tool SpaceX acquired last week. Once connected and authorized, an agent can view account information, analyze market data, place orders, and execute trades within limits the user sets up front. The platform also links agents to Binance Wallet’s agentic hub, the exchange’s x402 transaction verification API, and Binance’s Skill Hub.

The Guardrails — and the Gaps

Binance has built several constraints into the structure. Agents operate in dedicated subaccounts, isolated from a user’s main Binance account. The separation is meaningful: agents cannot initiate cryptocurrency withdrawals to external wallet addresses, and they cannot transfer assets from the primary account into the agentic subaccount — users must manually fund the subaccount before an agent can deploy capital. Binance’s VP of Product Jeff Li confirmed that withdrawals from subaccounts are disabled by default.

What Agent OS does not provide is a cap on how much an agent can lose within a subaccount. The exchange noted that it cannot observe an agent’s reasoning, information sources, or strategic logic — only the transactions it executes. The platform recommends users carefully review all specifications before final confirmation and restrict agent permissions to the minimum necessary, but the framework places significant responsibility on individual users to configure their own risk limits rather than imposing universal guardrails. TechCrunch’s coverage summarized it directly: keeping AI agents in check “is largely up to users.”

Part of a Fast-Moving Crypto AI Race

Binance is not moving alone. Coinbase launched Agentic Wallets in June 2026, enabling AI agents to hold funds and make autonomous on-chain transactions through non-custodial wallets. OKX debuted a decentralized AI marketplace in late June where autonomous agents can find tasks, fulfill contracts, and pay each other in stablecoins programmatically. Kraken deployed an AI investment advisor in July that suggests opportunities but requires manual trade approval. The four largest crypto exchanges have each introduced a different point on the AI autonomy spectrum within the past three months — from Kraken’s human-in-the-loop advisory model to Binance’s fully autonomous subaccount execution.

Binance co-founder Changpeng Zhao has been advocating for this convergence for months, publicly calling cryptocurrency the “native currency” of AI agents — an argument that crypto’s 24/7 markets, programmable infrastructure, and borderless settlement make it uniquely suited to autonomous AI operation compared to traditional financial markets with trading hours, settlement delays, and banking access requirements.

What It Means Bigger Picture

Binance’s Agent OS launch is the most operationally significant AI-in-finance development since algorithmic trading went mainstream, because it puts AI agency directly in the hands of retail participants rather than institutional quant desks. Someone with a Claude Code session and a funded Binance subaccount can now deploy an autonomous trading agent this afternoon. Whether that creates opportunity or accelerates losses for retail traders will depend heavily on how thoughtfully those users configure the permissions Agent OS hands them. Given the same week produced the first documented autonomous AI attack on a government’s infrastructure — covered in our story on Taiwan’s four-day AI agent breach — the question of what happens when AI agents operate without human checkpoints at every step is more than theoretical right now.

Sources: TechCrunch, Cryptopolitan, Crypto Economy

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Binance Just Let Claude and ChatGPT Trade Crypto on Your Behalf — No Human Required appeared first on Times Tabloid.

The Nevada Transportation Authority voted unanimously Thursday to hand Tesla, Waymo, and Uber the permits they need to operate paid commercial robotaxi services in Clark County — a decision that, if the fleets reach full scale, will put up to 8,000 driverless vehicles on Las Vegas roads within the next year.

Who Got What

Tesla received the largest allocation: permission to deploy up to 5,000 robotaxis across Clark County, with a provision allowing the company to expand statewide if it notifies the authority in advance. Waymo received a permit authorizing up to 1,000 autonomous vehicles in the region — the company had already been cleared to operate driverless vehicles in Las Vegas since July, but this new permit allows it to charge for rides for the first time. Uber received a permit for an additional 1,000 robotaxis, which it will operate through its autonomous vehicle partners: Hyundai subsidiary Motional, using Ioniq 5 crossovers, and Zoox, which brings its purpose-built all-electric autonomous vehicles. Zoox already held a separate permit covering 100 vehicles, adding to that total.

The combined ceiling of approximately 8,000 robotaxis would make Clark County one of the most heavily authorized autonomous vehicle markets in the world — though every company in the room was careful to separate what they’re allowed to deploy from what they expect to actually deploy.

The Gap Between Permission and Reality

Tesla’s Cybercab chief engineer Eric Early was explicit about it during the meeting: “The 5,000 has always been a ceiling for us. I don’t think we’ll be in a position by this time next year to deploy 5,000 vehicles, and it’s not the technology.” The company said it would be “extremely happy” to achieve half that number — 2,500 vehicles — by 2027. Tesla will begin with driverless Model Y electric crossovers rather than the purpose-built Cybercab, which is still in the ramp-up phase. Waymo has also been publicly cautious about its Las Vegas timeline, noting the city is on its expansion list without committing to a specific date or fleet size.

Uber’s approach differs from its competitors: the company has explicitly pitched a hybrid transportation model in which human drivers and autonomous vehicles operate within the same app and service infrastructure, allowing it to flex autonomous capacity up during peak demand without displacing its human driver base entirely — or at least, not all at once.

The Opposition That Was Heard

The permits didn’t pass without friction. Representatives from the Livery Operators Association and local taxi companies appeared at the NTA meeting to oppose the approvals, arguing the rollout is moving too fast and too far. The workforce concern is straightforward: Las Vegas is a labor-intensive city, and the taxi and ride-hailing workforce represents a significant group of workers for whom driverless vehicle competition is existential rather than incremental. Proponents argue the expansion will create a different category of jobs — maintenance, charging, cleaning, and fleet management — though skeptics note that those roles are fewer and pay differently from the driving jobs being replaced.

Why Las Vegas as the Test Case

Las Vegas has specific characteristics that make it attractive for large-scale robotaxi rollout: a dense tourist corridor with predictable, well-mapped routes, relatively mild weather compared to northern cities, and a state regulatory environment that has historically been receptive to autonomous vehicle experimentation. It also has an unusually high concentration of potential riders — tourists, convention attendees, and nightlife visitors who need transport but don’t have or want a rental car. The Strip corridor gives early fleets a natural operating zone before expansion into residential neighborhoods.

What to Watch

The combination of Tesla, Waymo, and Uber competing for the same riders in the same geography is a first — no other US city currently has three major robotaxi operators authorized to charge for service simultaneously. The real deployment pace will surface quickly, as each company will need to begin operations or face questions about why permitted capacity is sitting unused. Labor groups are already signaling they will push for state legislative action, meaning the regulatory picture may not be as settled as Thursday’s unanimous vote suggests.

Sources: TechCrunch, Engadget, InsideEVs

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Nevada Just Approved 8,000 Robotaxis for Las Vegas — Tesla Got Half of Them appeared first on Times Tabloid.

Security researchers at Israeli cybersecurity firm Dream have documented what experts are calling the first fully autonomous AI cyberattack against a government — a four-day operation against Taiwanese government systems in early July, suspected to be linked to China.

What the Attackers Built

The attack framework wasn’t a piece of sophisticated state-developed malware. It was assembled from publicly available tools. The operators used open-source AI agent frameworks — including Hermes, developed by Nous Research, and OpenClaw — to build a multi-agent system capable of mapping networks, discovering vulnerabilities, attempting intrusions, and adapting tactics automatically when one approach hit a wall. At peak activity, up to eight agents ran in parallel, each assigned to a different target or attack technique, operating more like a coordinated digital strike team than a script running a pre-programmed sequence.

The campaign unfolded between July 1 and July 4, 2026. Across twelve documented “attack waves,” the framework mapped 21 connected government systems, cracked 85 employee accounts, and exfiltrated at least 2,564 personnel records. Researchers retrieved a 160-megabyte archive of 1,395 files from the attackers’ operational workspace — including seven single sign-on client secrets, six internal database credentials spanning three different database systems, and internal network IP ranges. After cracking 85 accounts, the agents tested each one against connected internal systems through SSO; 84 of the 85 compromised credentials successfully accessed an internal information system because the SSO bridge trusted existing sessions without requiring additional authentication or multi-factor verification.

What Made It Autonomous in a Meaningful Sense

The system didn’t just execute pre-written instructions faster than a human could. It conducted what researchers called “learning cycles” — autonomous sessions in which the agents searched vulnerability databases, GitHub repositories, and security research publications to identify specific techniques and CVEs applicable to the targeted infrastructure, then adjusted their approach based on what they found. When one attack path was blocked, the agents reranked all available paths by probability of success using a Bayesian prioritization model and automatically switched to the next best option. The operation then expanded beyond initial targets to scan government IT supply chain vendors, a nuclear safety agency, a government email system, and at least seven energy companies — all in parallel.

Attribution and the Public Evidence

Dream’s researchers do not formally attribute the attack to the Chinese government or any specific hacking group, but note that internal operational documentation was written in simplified Chinese, pointing to a Chinese-language operator. Taiwan’s Ministry of Digital Affairs later confirmed an AI-assisted attack on government agencies in July. A spokesperson for China’s Ministry of Foreign Affairs told CNN it was not familiar with the situation.

Kenny Huang, chairman of the Taiwan Network Information Center, told CNN he believes this is the first disclosed case of a fully automated attack against a government. “It spells out one thing loudly,” Dream wrote in a post summarizing the findings: “the cost of running a competent attack has collapsed, but the cost of defending against one has not.”

Why This Changes the Threat Calculation

Prior AI-assisted cyberattack cases involved AI helping human operators write code, craft phishing emails, or search vulnerability databases. This operation reportedly required no ongoing human intervention during execution — once deployed, the agent framework pursued its objectives, learned, adapted, and expanded scope on its own across four days. That shift from AI-assisted to AI-autonomous attacks is exactly the scenario CISA and allied agencies have been warning about, particularly as the same week produced a federal emergency patch order for a critical vulnerability in the Ray AI compute framework widely used across the AI industry. See our coverage of that CISA advisory for the defensive side of the same dynamic.

What to Watch Next

The Taiwan incident is likely to accelerate calls for international norms around autonomous offensive AI systems, a governance gap that both the EU and US regulatory frameworks have not yet directly addressed. Whether similar operations are already underway against other governments — and have simply not yet been discovered or disclosed — remains an open and unsettling question.

Sources: Financial Times via Hackread, CNN, The Register

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Eight AI Agents Attacked Taiwan’s Government for Four Days — With No Human Pulling the Trigger appeared first on Times Tabloid.

If the numbers now circulating around Anthropic’s planned public offering are accurate, it won’t just be one of the largest AI listings in history — it could be the largest IPO ever recorded, full stop.

What Bloomberg Reported

Bloomberg reported on August 20 that Anthropic executives expect their IPO to match or exceed the size of SpaceX’s record-setting first share sale — a bar that stood at $75 billion at pricing and climbed to $86.2 billion once the overallotment option was exercised. The company is reportedly running numbers as it prepares to potentially file its public S-1 registration with the SEC as soon as the end of August, though discussions remain fluid. CFO Krishna Rao has sidestepped the question of valuation in recent investor briefings, and some backers are already pushing a $2 trillion figure — though no official target has been endorsed.

The Revenue Behind the Ambition

The financial trajectory Anthropic is leaning on is striking. Its annualized revenue run rate reportedly reached $65 billion by the end of July 2026, up from around $10 billion across all of 2025 — a roughly 600% increase in seven months. Second-quarter revenue came in at more than $11.5 billion, compared to $787 million in the same quarter a year earlier. Enterprise demand for Claude, particularly for coding agents and AI automation workflows, has been the primary engine: Anthropic overtook OpenAI as the world’s most valuable private company in May after closing a $65 billion Series H round that valued the company at $965 billion.

The Loss Figure That Will Dominate the S-1

None of that revenue growth changes the fact that Anthropic posted a net loss of approximately $42 billion in 2025 — up roughly fivefold from the prior year’s $8.3 billion loss — driven almost entirely by computing costs. Two specific contracts illustrate the scale of those costs: Anthropic has committed $100 billion to Amazon cloud services over ten years, and reportedly pays SpaceX approximately $1.25 billion per month for compute capacity through 2029. When Anthropic files publicly, those figures will appear in audited disclosures for the first time, alongside the full cost structure, share count, and governance terms that have been invisible to outside observers while the company remained private.

Supervoting Shares and the Race Against OpenAI

Anthropic is planning to issue supervoting shares, keeping control with CEO Dario Amodei and other co-founders even after the public offering — the same structure SpaceX used and a common choice for founders unwilling to cede governance to public shareholders. Morgan Stanley, Goldman Sachs, and JPMorgan are working on the offering, with additional banks potentially joining the roster. Notably, Anthropic’s CFO told investors there is no concern about OpenAI listing first, since OpenAI has now explicitly deferred its own IPO to 2027 — see our coverage of that timeline. Anthropic is widely expected to list ahead of its rival.

Why the Stakes Are This High

The IPO market context matters. US IPOs had raised $160.6 billion through August 19, already approaching the $195.2 billion record set in 2021. An Anthropic offering that clears SpaceX’s record could push 2026 to the highest annual US IPO volume in history almost on its own — a real-world test of how public investors, rather than venture backers, price AI companies whose revenue is doubling by the quarter but whose losses are doing the same. For years, frontier AI economics have been evaluated behind closed doors; the Anthropic S-1 will open them.

What to Watch Next

The most consequential moment will be the public S-1 filing itself, which will replace estimates and leaks with audited figures and legal disclosures — putting valuation debates on a factual footing for the first time. If the company files by month-end as reported, a roadshow could begin in September and pricing could follow shortly after.

Sources: Bloomberg via Japan Times, TechStartups, Benzinga

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Anthropic May Be Targeting the Biggest IPO in History — Here’s the Number Behind the Ambition appeared first on Times Tabloid.

For years, warnings about AI-powered attacks on critical infrastructure were largely theoretical. This week, the US government confirmed they are not.

The Joint Advisory

On August 19, CISA, the FBI, and the National Security Agency — along with several international partners — issued a joint cybersecurity advisory warning that foreign threat actors are actively targeting Siemens S7 Series programmable logic controllers, the industrial computers that automate physical processes at water treatment plants, energy facilities, chemical plants, and agriculture operations across the United States. The advisory is specific on the method: “The threat actors are conducting reconnaissance and capability development against US-based Siemens PLC installations using AI-generated exploitation scripts disguised as legitimate monitoring tools.” Officials were blunt about the current state: “This is not a theoretical risk — it is an active threat.”

What Has Already Happened

The advisory follows a wave of actual incidents. Cyberattacks disrupted more than 30 community water systems in Minnesota in attacks that began last Sunday and continued through Monday. Across the US, attacks have reached at least 12 states, with CISA confirming that compromises have “resulted in boil water notices and sustained manual operations.” The attackers’ playbook focuses on operational disruption rather than data theft: once inside exposed PLCs, they modify passwords to lock out operators, change IP addresses to disconnect devices from the network, and alter automated settings that force utilities to switch to slower manual control.

CISA Acting Director Nick Andersen told Recorded Future News directly: “CISA is currently observing a significant increase in cyber threat actors targeting programmable logic controllers at water utilities,” adding that “threat actors are targeting water entities of all sizes” — meaning this isn’t a problem confined to large metropolitan systems with complex infrastructure and sophisticated IT teams. Small rural water utilities, often running older equipment with minimal cybersecurity staffing, are equally in scope.

The AI Angle That Changes the Threat Profile

What makes this advisory distinct from earlier infrastructure warnings is the explicit role AI is playing on the attackers’ side. Officials describe adversaries using AI-generated scripts to scan the internet for vulnerable Siemens devices and develop exploitation code faster than traditional manual methods allow. That capability shift dramatically compresses the timeline between a device being exposed and a successful compromise — the same dynamic Wiz’s autonomous Red Agent demonstrated from the defensive side when it found and exploited a Snowflake vulnerability within five days of its creation; see our coverage of that incident.

US officials are investigating whether Iran is behind the current water infrastructure attacks — Iranian state-linked hackers, operating under personas including CyberAv3ngers, have targeted US water systems in multiple prior campaigns stretching back to 2023. The current advisory stops short of formal attribution, but the operational signatures are consistent with that history.

What CISA Is Telling Operators to Do Right Now

The advisory’s recommendations are specific: take inventory of all Siemens S7 Series PLCs, install available security patches immediately, verify that no controllers are reachable from the public internet, strengthen access controls and authentication requirements, and monitor for anomalous activity. CISA’s underlying message is that the fundamental exposure — critical industrial control systems accessible from the internet — is the root problem that patches alone can’t fully address. “CISA urges critical infrastructure owners, operators, and integrators to remove publicly exposed PLCs and other operational technology from the internet as soon as possible,” a separate earlier advisory stated.

The Bigger Picture

Water infrastructure cybersecurity has been a known and documented vulnerability for years, yet a meaningful share of US water utility PLCs remain internet-exposed — as CISA’s own scans have continued to confirm even as attacks mount. The combination of aging operational technology, limited cybersecurity budgets at small utilities, and now AI-accelerated attack tooling has narrowed the window for catching up considerably. Today’s patch deadline for the Ray AI framework vulnerability — covered in our companion story on CVE-2025-62593 — is a reminder that the same compressed remediation timelines now apply across the full spectrum of AI-adjacent infrastructure, from model training clusters to the water utilities those systems are increasingly used to attack.

Sources: TechCrunch, Newsweek, Recorded Future News / The Record

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Hackers Are Using AI Scripts to Attack US Water Systems — And 12 States Have Already Been Hit appeared first on Times Tabloid.

The open-source software that sits at the heart of AI workloads at Amazon, Apple, and OpenAI has a critical, actively exploited vulnerability — and federal agencies have until today, August 20, to patch it or take affected systems offline.

What Ray Is and Why This Matters

Ray is an open-source distributed computing framework built by Anyscale, designed to take Python workloads — machine learning training runs, model inference pipelines, data processing jobs — and scale them across clusters of CPUs and GPUs. With more than 43,000 GitHub stars and adoption at some of the largest AI companies in the world, it’s not a niche tool. When Ray has a critical vulnerability, the blast radius is potentially enormous: affected systems typically aren’t individual laptops but pools of expensive compute holding proprietary models, training data, and cloud credentials.

The Vulnerability

CVE-2025-62593 is a code injection flaw affecting Ray versions prior to 2.52.0, rated 9.4 out of 10 under CVSS v4 — effectively the highest severity tier. CISA Adding it to its Known Exploited Vulnerabilities catalog on August 17, and gave US Federal Civilian Executive Branch agencies three days to remediate, under the agency’s Binding Operational Directive 26-04, which applies its shortest window to vulnerabilities granting total control of an exposed asset after exploitation.

The attack mechanism is technically precise. Ray exposes HTTP API endpoints — including /api/jobs and /api/job_agent/jobs/ — that are designed to submit and execute code on a cluster. Ray’s intended defense against a malicious website reaching a locally running Ray instance was to check whether the HTTP User-Agent header began with “Mozilla,” on the logic that browser requests carry that prefix. The problem: Firefox and Safari both allow scripts using the Fetch API to modify that header. Combined with a DNS rebinding attack — which tricks a browser into connecting to a local address after initially resolving to a benign external one — an attacker can reach the Ray API through a developer’s browser and execute arbitrary code simply by persuading that developer to visit a malicious site or see a malicious advertisement.

Already Actively Exploited in the Wild

CISA’s KEV listing reflects confirmed active exploitation, not just theoretical risk. Security firm Bitsight linked the vulnerability to RondoDox, a DDoS botnet known for aggressively scanning the internet for exploitable endpoints, which began incorporating a proof-of-concept for this flaw as early as November 24, 2025 — two days before the CVE was formally published. Researchers at Oligo Security separately tracked a campaign called ShadowRay 2.0 that specifically targets AI infrastructure clusters equipped with high-end Nvidia GPUs, compromising them for cryptomining and other purposes. The flaw was first disclosed in November 2025; its trajectory from disclosure to active exploitation to federal emergency patch order took less than nine months.

The Fix and What To Do

The vulnerability is patched in Ray 2.52.0. For any organization running Ray in a development or production environment, the immediate action is straightforward: identify any Ray installation older than 2.52.0 and update it, with particular urgency on machines that run Ray while also browsing the web in Firefox or Safari. CISA additionally recommends isolating Ray deployments from internet-accessible surfaces wherever possible — a reminder that Ray was architected as a trusted internal compute framework, not a public-facing service, and exposing its API endpoints to the internet was never an intended configuration.

Part of a Bigger Pattern

This incident joins a growing catalogue of AI infrastructure vulnerabilities disclosed in 2026. Where earlier security stories this year focused on AI agents escaping their sandboxes or AI coding tools introducing flaws into production code, this one sits a layer deeper: the compute framework itself, used to run the models rather than built by the models, being actively exploited to steal the GPU clusters that make large-scale AI possible. It’s a reminder that securing AI systems means securing not just the models and the products built on top of them, but the entire infrastructure stack underneath — a point covered from a different angle in our reporting on the Wiz Red Agent finding exploits in Snowflake’s pipeline.

Sources: CISA KEV Catalog, The Hacker News, The Register

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post The AI Framework That Powers Amazon, Apple, and OpenAI Has a Critical Flaw — And the Patch Deadline Is Today appeared first on Times Tabloid.

SpaceX completed its $60 billion acquisition of AI coding startup Cursor on August 14-15, finalizing what is reported to be the largest acquisition of a venture-backed startup in history and folding one of the most widely used AI development tools directly into Elon Musk’s expanding tech empire.

How This Deal Came Together

The acquisition followed an unusually fast trajectory. In April 2026, SpaceX and Cursor — built by startup Anysphere — announced a compute partnership giving SpaceX the option to buy Cursor for $60 billion. In June, SpaceX went public in what was itself one of the largest IPOs in history, raising $75 billion at $135 per share. Four days after its IPO, SpaceX formally announced it was exercising the Cursor acquisition option. The all-stock deal then closed in mid-August following regulatory review. SpaceX had already absorbed Elon Musk’s AI company xAI earlier in the year, rebranding the combined entity as SpaceXAI. Cursor now operates under that umbrella.

What Cursor Gets: The World’s Largest GPU Fleet

Cursor’s announcement repeatedly emphasized a single point: it now has access to “the largest fleet of GPUs in the world.” SpaceX’s Colossus supercomputer — built around hundreds of thousands of Nvidia GPUs — is the specific infrastructure Cursor is pointing to. The practical implication is that Cursor can now train larger, faster, and more capable models than any independent AI coding startup could plausibly afford, and can offer those models to its customers at lower cost than running on rented cloud compute from Amazon, Google, or Microsoft.

The collaboration was already producing results before the deal closed. Cursor worked with SpaceXAI on the development and training of Grok 4.5, their first jointly built model, and then Grok 4.6 — designed to “excel at coding, agentic tasks, and knowledge work” and deliver results at lower cost. Those releases gave developers a preview of what the combined entity can produce when Cursor’s product expertise is paired with SpaceXAI’s compute and model research.

The Competitive Question Developers Are Already Asking

Cursor’s popularity has been built partly on flexibility: it supports multiple underlying AI models, including Anthropic’s Claude and OpenAI’s GPT series, letting developers pick the model that works best for their workflow. That flexibility is now structurally at risk. SpaceX has a clear financial incentive to shift workloads toward its own Grok models rather than paying Anthropic or OpenAI for every Claude or GPT API call routed through Cursor. Notably, SpaceX has been renting its Colossus compute to Anthropic and Google as customers — a relationship that becomes somewhat more complicated now that SpaceX owns the coding tool that competes directly with tools both companies power.

Enterprise market share data from Ramp spending suggests Cursor’s dominance is being contested: its share slipped from 41% in June 2025 to 26% by May 2026, even as revenue climbed to roughly $2.6 billion annualized by June — suggesting GitHub Copilot and other rivals are closing the gap even as the overall market grows. The SpaceX acquisition looks less like a bet on a comfortable incumbent and more like a move to lock in compute advantages before the competitive window narrows further.

The Regulatory Attention This Will Draw

The FTC is expected to scrutinize the deal under vertical integration frameworks — a company that owns frontier AI models, the world’s largest GPU fleet, and now the dominant enterprise coding tool raises obvious questions about whether competitors can access infrastructure on fair terms. Rivals have already raised concerns about GPU access disparity. Whether regulators move beyond scrutiny to action remains to be seen, but the deal arrives at a moment when antitrust attention on AI infrastructure concentration is at its highest point in years.

Why This Fits the Broader Moment

This acquisition is one data point in a rapidly concentrating AI industry where compute access is becoming the defining competitive moat. It’s directly analogous to Google’s $12.2 billion chip deal with Marvell, also announced this week — see our coverage of that deal — both transactions showing how major players are racing to vertically integrate their AI stacks from hardware up through developer tools that sit directly in front of millions of engineers.

Sources: TechCrunch, Engadget, SatellitePro ME

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post SpaceX Just Closed the Biggest AI Startup Acquisition Ever — And Cursor Now Has the World’s Largest GPU Fleet appeared first on Times Tabloid.

When Marvell Technology disclosed a new custom chip partnership with Google on August 19, its stock jumped more than 11%. Broadcom’s fell more than 5%. That divergence tells you almost everything about what the deal actually means.

What Was Announced

Marvell filed a regulatory disclosure revealing it has issued Alphabet’s Google a warrant to purchase up to 58,970,907 shares of its common stock at $206.58 per share — worth approximately $12.18 billion if fully exercised. The underlying commercial agreement was signed on July 29, and it covers a broad range of custom silicon attached to Google’s tensor processing unit ecosystem, including AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory computing technology. If fully exercised, the stake would make Google Marvell’s fifth-largest investor. The warrant is exercisable until August 18, 2033.

The shares don’t vest all at once. Only about 1.4 million vest automatically in the first year. The remaining 57.6 million are tied directly to how much Google spends with Marvell — approximately 240,000 shares vest for every $500 million in qualifying chip purchases. To earn the full $12.18 billion in equity, Google would need to send roughly $120 billion in chip revenue Marvell’s way through fiscal 2033. That $120 billion is a vesting threshold, not a spending commitment, but it signals the potential scale of the relationship if Google follows through.

Why Broadcom Sold Off

Broadcom has been Google’s primary custom chip partner, holding a deal that runs through 2031 and covers future generations of Google’s most advanced silicon. The Marvell announcement doesn’t replace that — Broadcom remains engaged — but it signals that Google is deliberately building a second, parallel custom chip relationship rather than deepening its dependence on a single supplier. For investors, that diversification reads as a relative vote of no-confidence in Broadcom’s exclusive positioning, even if the absolute relationship is unchanged. Marvell and Broadcom don’t compete on identical products; Morningstar analyst William Kerwin called the Marvell deal “a big win” while noting it likely reflects a “growing pie at Google for new sources, rather than” a direct transfer of business away from Broadcom.

Part of a Pattern Across Big Tech

The deal fits a template becoming standard among the largest AI infrastructure spenders: take a meaningful equity position in a key chip supplier in exchange for committing significant procurement volume, structuring the stake to scale automatically with spending. In October 2025, AMD struck an analogous deal with OpenAI — chips worth tens of billions in potential annual revenue, with OpenAI holding an option to buy up to roughly 10% of AMD. The structure aligns incentives both ways: the chip company gets a large committed customer; the customer gets equity upside on its own spend and a deeper engineering partnership with a supplier that has strong financial reasons to prioritize their roadmap.

The Broader Race to Build Custom Silicon

Demand for custom, task-specific chips has surged as AI companies discover that general-purpose Nvidia GPUs — however powerful — aren’t optimally efficient for inference, the process of running trained models at scale rather than training them in the first place. Google’s TPUs are the most mature example of this category, and the Marvell deal specifically targets the hardware that surrounds and enables TPUs rather than the TPUs themselves — the storage, networking, and memory controller chips that determine how efficiently a TPU cluster can actually be used in production. This connects directly to the same chip demand story driving TSMC’s blowout quarter and Samsung’s price hikes; see our coverage of Samsung’s foundry price increases for how AI demand is squeezing every layer of the hardware supply chain.

What to Watch Next

Marvell reports earnings August 27, and analysts are watching for detail on the revenue ramp tied to the Google deal. Prediction market data puts Google at just 7% odds to hold the best AI model by end of 2026, compared to 66% for Anthropic and 14.5% for xAI — a sobering context for a deal that’s ultimately a bet on winning the AI infrastructure race if not the model race outright.

Sources: Reuters via Yahoo Finance, Bloomberg, CNBC

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post Google Just Gave Marvell a $12.2 Billion Vote of Confidence — And Broadcom Lost 5% the Same Day appeared first on Times Tabloid.

For a company that could be heading for one of the largest stock market debuts in US history, OpenAI’s chief financial officer is doing her best to make it sound like a routine Tuesday.

What Friar Told Employees

On August 19, OpenAI CFO Sarah Friar addressed employees at an all-hands meeting and delivered the clearest public timeline yet for the company’s IPO: the company “will be a public company in 2027,” or sooner if “our business continues to inflect.” She immediately tried to lower the temperature around the announcement, telling staff: “The IPO is not a finish line, it is a milestone, another fundraise. We raised $122 billion in March, and that gives us flexibility.” The remarks were reported by CNBC, citing two people familiar with the meeting who were not authorised to speak publicly.

Friar also addressed the prospect of rival Anthropic going public first. She told employees Anthropic’s own confidential SEC filing could become public within weeks — potentially setting up an Anthropic listing as early as September — and that OpenAI is unbothered: “We are running our own race.” Both companies have separately filed confidential S-1 prospectuses with the SEC, and whichever lists first will effectively set the public market benchmark the other gets measured against. See our earlier coverage of OpenAI’s IPO plans and its financial profile for the full background on what investors will actually see in the filings.

The Revenue Slides Behind the Reassurance

Friar backed her timeline with unpublished numbers shown on slides during the meeting. OpenAI’s revenue run rate is up 35% quarter-to-date in the current quarter, its enterprise revenue run rate is up 50%, and its AI coding and work product has reached 20 million weekly active users. OpenAI separately told investors it generated $6.7 billion in revenue in the second quarter — up 18% from the first quarter — though the Wall Street Journal described some of the Q2 figures as “tepid” and noted the company’s ongoing operating losses remain substantial at an estimated $14 billion for the full year.

OpenAI’s Codex enterprise coding product alone has hit $20 million in revenue, a notable ramp that coincides with the company’s newest model releases this summer. Enterprise services now account for more than 40% of total revenue and are growing faster than the consumer segment — a structural shift that matters enormously for how public investors will value the business, since enterprise contracts tend to be stickier and more predictable than consumer subscriptions.

The Executive Turnover Underneath the Optimism

Friar’s upbeat presentation lands amid a wave of senior departures that has prompted questions from some of OpenAI’s financial backers. Revenue chief Denise Dresser departed last week just eight months into the job — following former COO Brad Lightcap’s exit the week before after eight years at the company. Friar joins those two as a relatively recent hire herself, appointed to the CFO role earlier this year after serving as CEO of Nextdoor and CFO at Square. OpenAI President Greg Brockman told CNBC the turnover rate is “not actually that atypical” and attributed the attention to the company’s high public profile rather than organizational dysfunction.

The $7 Billion Secondary Sale Running in Parallel

Separate from the IPO, OpenAI completed a $7 billion secondary share sale this week, giving current and former employees the ability to sell stock at the company’s $852 billion private valuation ahead of any public offering. That mechanism both relieves employee liquidity pressure and signals that the company can support a market for its stock even before the S-1 becomes public — another data point Friar can point to when arguing the IPO is simply the next step rather than an existential milestone.

What to Watch Next

The central tension heading into the eventual IPO is the same one Friar’s presentation didn’t fully resolve: whether revenue momentum at the 35% and 50% quarterly growth rates she cited can outrun a cost structure that currently loses roughly $14 billion a year. Public markets will demand audited answers to that question in a way that private fundraising rounds haven’t.

Sources: CNBC, QZ, Cryptopolitan

Disclaimer: This content is meant to inform and should not be considered financial advice. The views expressed in this article may include the author’s personal opinions and do not represent Times Tabloid’s opinion. Readers are advised to conduct thorough research before making any investment decisions. Any action taken by the reader is strictly at their own risk. Times Tabloid is not responsible for any financial losses.

The post OpenAI’s CFO Told Employees the IPO Isn’t a Big Deal — Then Revealed a 35% Revenue Jump appeared first on Times Tabloid.