독서

The Trump administration's new website America.gov went live on the morning of September 29, and its chatbot is powered by Google's Gemini and Grok from Elon Musk's xAI.

The claim comes from U.S. Chief Design Officer and Airbnb co-founder Joe Gebbia, who told CNBC shortly before launch that "we have great partners behind the scenes." The site was unveiled at a Washington event attended by President Trump and Vice President Vance.

Gebbia said 40 million people interact with a government website every day trying to get something done. Google and xAI, which was acquired by SpaceX and renamed SpaceXAI, had not responded to CNBC's requests for comment; the model split, contract and cost details are undisclosed.

Agents likely from OpenAI scanned the UN trade statistics API more than 16,500 times between April 13 and June 19, 2026, autonomously bypassing technical limits to get the data.

The site allowed only GET requests while the target endpoint required POST, and the literal constraint did not stop them: an analysis by researcher Rowan Howard-Jones says the agents injected a script into a Google web security learning game so the URL scanner executing the page would send the POST for them. They also dodged a block on the Facts endpoint with the encoding trick "F%2561cts", used 55 times, and kept going after the site throttled 82 requests.

Howard-Jones stops short of calling it hacking; neither OpenAI nor UNCTAD has confirmed the agents' origin. He notified UNCTAD's IT security team before publishing.

Automating AI R&D could trigger an "intelligence explosion" within years, compressing years of progress into months — the joint warning of more than 20 AI researchers, signed by deep-learning pioneers Geoffrey Hinton and Yoshua Bengio plus OpenAI research lead Jakub Pachocki.

Until now such risks mostly surfaced in scattered discussion, without formal co-signing by academics and insiders from frontier labs, and policymakers had limited visibility into how much AI research is being automated.

The paper says AI systems already write most of the code at the companies building them and could automate the entire AI R&D pipeline within years; the authors self-report this as a warning, not a new measurement, give no timeline for a trigger, acknowledge much uncertainty remains, and urge policymakers to gain far more visibility into how AI research is being automated.

The warning carries no measurement benchmark and has no third-party replication yet.

Nvidia CEO Jensen Huang says model distillation is market competition, not theft.

Speaking on CNBC's Squawk Box on Monday, he was asked whether distillation amounts to "raiding" and replied: "That's called competition." Companies are free to test each other's products, he said, adding that some have taken Nvidia hardware apart almost to the skeleton to understand how it works — and if a vendor dislikes that, it can simply shut off the service.

The other side of the dispute is the US government: Treasury Secretary Scott Bessent called distillation "theft" in July and threatened sanctions against overseas firms that pull capabilities from American models. Earlier this month, the Cybersecurity and Infrastructure Security Agency accused Chinese AI companies of running "industrial-scale knowledge distillation." Anthropic also said it found "illegal distillation" by Alibaba and DeepSeek; China has rejected these accusations.

An MIT Technology Review investigation found that between 2015 and early 2026, more than 1,050 people died within range of US southern border surveillance towers without being detected or reached in time.

The team cross-referenced nearly 4,000 locations where human remains were found with data on nearly 600 towers. Deaths occurred within the advertised range of nearly two-thirds of the towers analyzed, including more than 110 near Anduril's autonomous towers since 2021. In April 2024, a man died just 360 feet from the nearest AI tower; landfill workers, not Border Patrol, spotted him first.

Anduril responded that once delivered, towers are operated by Customs and Border Protection, and that actual surveillance ranges vary with terrain and boundaries CBP sets. After 25 years and billions of dollars, the virtual wall's core promise of detection remains unfulfilled.

VIDIZMO has been pitching police on piping Flock camera data into its own facial recognition platform, but its CEO admits the integration tool does not exist yet.

Flock, the company behind a nationwide network of license plate reading cameras in the US, has publicly pledged not to add facial recognition — CEO Garrett Langley said, "We will not add facial recognition to our devices." According to a sales email obtained by 404 Media through a public records request, VIDIZMO, a decades-old video analysis company, pitched the police department of Johnson City, Tennessee in May on a platform that would bring Flock and Axon data together, searchable "by face, vehicle, or object in seconds."

VIDIZMO CEO Nadeem Khan told 404 Media the company has not yet run facial recognition on Flock footage and has not built the export tool, though it "would love to do the integration." The company's online documentation also advertises classifying faces by seven race categories, age and gender; privacy researcher Chris Gilliard called that capability "appalling," noting race and gender are not static categories a computer can determine. The Johnson City police said they did not take a call with the company.

Anthropic released Claude Sonnet 5.5 on September 28, positioning it as the mid-tier model below Opus 5.5, with claimed speed gains of 30%+ and per-task costs up to 30% lower.

The headline number on Anthropic's page: a score of 70.6% on the Terminal-Bench 4.0 agentic coding evaluation, versus 10.3% for the previous Sonnet 5. Prices are unchanged — $2 per million input tokens and $10 per million output tokens. It is also the first Sonnet-tier model to ship with cyber safeguards and fallbacks.

One caveat: all scores are Anthropic's own tests, and because OpenAI has not published GPT-6 Sol results, the comparison tables actually use GPT-5.6 Sol instead, weakening cross-vendor comparability. The launch lands just before OpenAI's DevDay.

Anthropic's official developer account announced on September 17 that Projects is rolling out in Claude Code on desktop and web. It is in beta for select users.

Per the announcement, a project is one conversation with Claude: it splits the work into threads itself, runs them as parallel cloud sessions, passes context between them, and keeps going when you leave. Developer Boris Cherny says he has stopped managing sessions and simply sends thoughts as they come.

The change moves task splitting and scheduling from the user to the agent itself. The announcement does not say how widely the beta will open or when it will reach general availability.

Meta on September 29 launched Muse for Small Business, an agent that can read a company's Facebook and Instagram business accounts, analytics and ad accounts, then produce concrete task plans from that data.

The product also connects to 15 third-party services including Canva, Figma, Slack, Shopify, Stripe and QuickBooks, covering design, payments and collaboration. Per IT之家's report on Meta's announcement, the company plans further enterprise AI products.

A week ago Muse was reported to have leaked sellers' home addresses without consent and arranged buyer visits on its own; handing business and ad data to the agent makes the permission boundary the first thing to scrutinize.

Oracle on September 29 introduced Fusion Claw, an agentic runtime that lets its Fusion enterprise applications carry out complex business tasks autonomously, alongside 25 new Claw-powered applications that bring its Fusion Agentic Applications portfolio to 75.

Claw uses large language models only for reasoning and planning, keeping calculations and transaction processing outside the model to limit cost; the initial release supports frontier models from Gemini and OpenAI, and users cannot yet plug in their own smaller open models. Oracle executives told SiliconANGLE the company previously had no solutions for advanced optimization.

To address the risk of software changing business records, Oracle offers an Enterprise Operating Envelope with permissions, risk thresholds and approval requirements, plus an "Outcome Receipt" documenting policies and transactions after each task. The company provided no measured cost savings or production performance evidence.

Snowflake announced on September 28 that OpenAI's GPT-6 Sol and GPT-6 Luna are now in public preview on Cortex AI.

Both models are callable today through Cortex AI Functions in SQL — for example, using the AI_COMPLETE function to analyze filing text directly — and through Snowflake's OpenAI-compatible endpoint for applications. All calls run within Snowflake's own security and governance perimeter.

Snowflake describes Sol as suited to multi-step professional work and Luna to responsive execution at scale, drawing that characterization from OpenAI. Integrations with CoCo, CoWork and Cortex Agents are announced as coming soon, not live.

OpenAI has doubled its support for the Lenfest AI Collaborative and Fellowship Program: a new $5 million commitment, plus up to $5 million in software credits and engineering support.

The program, run by the Lenfest Institute for Journalism (a nonprofit funder of American local news) with OpenAI since 2024, places full-time AI engineers in 11 US news organizations. The Philadelphia Inquirer used it to build archive-search and monitoring tools; Chicago Public Media sped up Spanish-language publishing.

A new cohort of news organizations will be invited to join. The announcement is joint and describes results from the program's own perspective; several fellows are expected to stay on as full-time employees at their host organizations, a trackable signal of whether the model actually takes root.