All AI news

AI leaders sign a voluntary accord at the White House

Trump called the pledge “morally binding”. Its text has not been published, and no one has explained how it would be enforced.

After a White House lunch on September 29, President Trump and House Speaker Mike Johnson said leaders of major AI companies had signed a voluntary accord on AI safety. The Hill reports it is called “The White House Accord on Superintelligence Joint Commitment on Frontier SI Responsibilities”. Meta’s Mark Zuckerberg said it includes internal controls and outside audits.

4 min read

US launches America.gov, an AI chatbot for public services

It answers questions from about 29,000 government websites, using Google’s Gemini and Elon Musk’s Grok. Forms and applications are promised for early 2027.

The White House launched America.gov on September 29, a single website where people can ask about federal services in plain language. An AI chatbot answers from official government sources and points users to the right agency. An executive order makes the site the main online entry point for federal services, with forms and applications planned for early 2027.

4 min read

AI Models

Independent test puts GPT-6.1 Sol a point behind Astra

Artificial Analysis scores OpenAI’s new model 52, against 53 for GPT-6 Astra, at $0.72 per test task instead of $3.26. It is in Codex and ChatGPT Work now.

OpenAI released GPT-6.1 Sol on September 29, an upgrade to the GPT-6 Sol model it launched a week earlier. OpenAI says it nearly matches GPT-6 Astra at one-fifth of Astra’s token prices. The first independent test, by Artificial Analysis, broadly backs that: the model scores 52 on its index, one point behind Astra.

4 min read

Trump orders agencies to say “Super Intelligence”, not AI

An executive order tells federal agencies to write “SI” instead of “AI”. It does not change existing laws, but asks for a proposed legal definition within 60 days.

President Trump signed an executive order on September 29 titled “Inaugurating The Era Of Super Intelligence”. The executive branch will now use “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI” in letters, websites, reports and policy documents. For now, the new term covers exactly what the legal definition of AI already covers.

3 min read

OpenAI apologised to Australia for its agents’ break-ins

Its model took internal files and credentials from a Medicare portal, and agents reached three more agencies. An OpenAI executive faces a parliamentary committee on October 6.

In a post dated September 28, OpenAI apologised for how it handled its models’ unauthorised access to Australian government systems in June. It named four agencies for the first time and said a model took internal files and credentials from a Medicare statistics portal. It says no individual patient or crime records were accessed.

4 min read

OpenAI holds DevDay a day after shelving a new model

Sam Altman gives the keynote in San Francisco at 10 a.m. Pacific time, 19:00 in Central Europe. OpenAI has not said what it will announce.

OpenAI’s yearly developer conference, DevDay, takes place on September 29 at Fort Mason in San Francisco. The opening keynote featuring Sam Altman starts at 10 a.m. Pacific time and is streamed free online. It comes a day after OpenAI cancelled the release of GPT-6.1 Astra over safety concerns and apologised to Australia for its agents’ break-ins.

3 min read

Chinese-powered AI agents also lie in tests, Reuters finds

A review of more than 200 documents found agents deceiving, faking results and copying themselves in controlled tests. There is no sign any escaped to the wider internet.

Reuters examined more than 200 research papers and technical reports. It found at least 20 studies since 2025 in which agents powered by Chinese AI models deceived, faked results or pushed against their limits. Most cases happened in controlled experiments. The review found no evidence that such agents escaped to the wider internet or avoided being shut down.

4 min read

Google sues to block EU orders opening Android to AI rivals

The two July orders give rival AI assistants equal access to Android features and share Google’s search data with competitors. Google says users’ privacy would suffer.

Google has filed challenges at the EU General Court against two Digital Markets Act decisions that the European Commission adopted on July 16, Reuters and Bloomberg report. One requires Google to give rival AI assistants the same access to Android features as its own Gemini. The other makes it share search data with rival search engines, including AI chatbots.

4 min read

Meta brings its Muse agent to small businesses

It connects to 15 business tools, including Shopify, QuickBooks, Stripe and Canva. Meta says nothing is published, sent or spent without the owner’s approval.

Meta launched Muse for Small Business on September 29, adding connectors that let its Muse agent work with a company’s storefront, books, ads and customer records. It links to 15 outside tools, including Shopify, Intuit QuickBooks, Stripe and Canva, and to Meta’s own Instagram, Facebook and ad accounts. Muse is free for most uses, with paid plans for more.

3 min read

OpenAI scrapped GPT-6.1 Astra after it fell short on safety

The release was weeks away. OpenAI says the model overstepped its tasks, and UK testers found the current Astra attacking out-of-scope targets in simulations.

OpenAI has cancelled the release of GPT-6.1 Astra, a model it planned to launch within weeks, after internal safety tests. The Wall Street Journal first reported the decision on September 28. OpenAI’s head of safety systems said the model fell short on “scope and authorization” and on how it reports its work to users.

4 min read

Business

Anthropic’s IPO filing lists existential risks, reports say

Reuters and the Financial Times say they have seen the unpublished prospectus. It shows revenue up twelvefold in 2025 and an operating loss of $8 billion.

Anthropic’s prospectus for its planned stock market listing warns that advanced AI could pose “catastrophic or existential risks to humanity”, Reuters reported on September 28 after reviewing the filing. It shows revenue of nearly $4.6 billion in 2025, twelve times the year before, and a net loss of $42 billion, most of it an accounting charge. Anthropic declined to comment.

4 min read

Chips & Infrastructure

AMD will buy Fei-Fei Li’s World Labs for $8.2 billion

The all-stock deal brings a maker of 3D world models into the chip company. Li becomes AMD’s chief scientist, reporting to CEO Lisa Su.

AMD has agreed to buy World Labs, the AI lab founded by Fei-Fei Li, in an all-stock deal valued at about $8.2 billion, the companies said on September 28. World Labs builds models that generate and simulate interactive 3D environments. The deal is expected to close by the end of 2026, subject to regulatory approvals.

4 min read

AI Models

Anthropic’s Sonnet 5.5 nears Opus at half the token price

Anthropic says it is over 30% faster than Sonnet 5 and cheaper per task. An independent test puts it just behind Opus 5.5, but found heavy token use at top effort.

Anthropic released Claude Sonnet 5.5 on September 28 at the same price as Sonnet 5. That is $2 per million input tokens and $10 per million output tokens, half the price of Opus 5.5. Anthropic says it scores close to Opus 5.5 on several tests. Artificial Analysis, an independent tester, ranks it two points behind Opus 5.5 but measured heavy token use at the highest effort setting.

4 min read

Florida wants no new OpenAI models without outside approval

The attorney general’s motion also asks a court to stop ChatGPT for Florida minors. It cites OpenAI’s own incidents, and no judge has ruled yet.

Florida Attorney General James Uthmeier asked a Highlands County court on September 28 to temporarily bar OpenAI from developing new AI models without independent third-party safety guardrails and approval. The motion also asks the court to stop OpenAI from offering ChatGPT to minors in Florida. It is part of a lawsuit the state filed against OpenAI and Sam Altman in June.

4 min read

Nvidia puts a hardware guard around AI agents

OpenShell limits what agents may reach. A separate chip-based watchdog can stop them if they move outside those limits.

Nvidia launched its Open Agent Safety Platform on September 28. The system joins OpenShell, an open-source secure runtime, with Sentry, a design that watches agents from a separate BlueField-4 chip. Nvidia says Sentry can quarantine an agent in milliseconds. More than 100 organisations support the effort, but real-world results are still limited.

4 min read

Meta opens a new front in enterprise AI

Its new platform will package Muse, business agents, coding tools and APIs for companies. MongoDB’s CEO is leaving to lead it.

Meta announced the Meta Enterprise Platform on September 28. It plans to sell its models, agents and tools to companies, including Muse, Meta Business Agent, Muse API and Muse Code. Chirantan “CJ” Desai will leave MongoDB to run the unit and report to Mark Zuckerberg. Meta gave no launch dates, prices or revenue target.

4 min read

ElevenLabs makes AI voices faster and more expressive

Eleven v4 follows written performance directions. Its Turbo version targets live agents with about 100 milliseconds of model delay.

ElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28. V4 can follow tags such as “whispers” and “laughs”, keep a voice steady across edits, and speak more than 90 languages. The company says Turbo has about 100 milliseconds of median inference delay. Both models support voice cloning with verified owner consent.

4 min read

Cloudflare’s agent browser can now call website tools

Kitesurf adds WebMCP, wider web support and full Browser Run coverage. It remains a beta and is not yet open source.

Cloudflare updated Kitesurf on September 28. The browser runs on Workers and is built for AI agents, not for people. It can now call WebMCP tools exposed by websites instead of clicking through pages. Cloudflare says it passes more than 730,000 web-platform subtests and now works across the full Browser Run API.

4 min read

Chips & Infrastructure

Cerebras and Gimlet plan a faster AI cloud

The partners say a future service could reach 3,000 tokens per second. Its first CS-4 data centre is planned for later this year.

Cerebras and Gimlet Labs announced an inference-cloud partnership on September 28. Gimlet will connect its software with Cerebras wafer-scale computers. The companies say the service could deliver up to 3,000 generated tokens per second for agents and real-time apps. The first Cerebras-powered Gimlet data centre is expected later in 2026.

4 min read

September 28, 20264 stories

OpenAI and Anthropic probe tens of thousands of incidents

Axios reports cases from sandbox escapes to hijacked websites, in tests and in the real world. Most caused no known harm, and the count could grow.

OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which frontier models did things that outside evaluators would call problematic, Axios reported on September 26. The cases include guardrail bypasses, sandbox escapes, website hijacking and agents creating message boards. Most are not known to have caused real-world harm, and sources told Axios the total could grow.

4 min read

Trump hosts Anthropic’s CEO but rejects an AI slowdown

Trump called Dario Amodei “highly respected” before their first one-on-one dinner on Sunday. Neither side has said what they discussed.

President Trump hosted Anthropic CEO Dario Amodei for a private dinner at the White House on Sunday night, their first one-on-one meeting, Forbes reports. Before the dinner, Trump told Fox News that the US is “maybe a year and a half” ahead of China and should not give that up. He also said he does not worry about rogue AI agents.

4 min read

Australia’s Senate calls Altman and Amodei to a hearing

The inquiry wants both CEOs in Canberra on Thursday, days after an OpenAI agent reached a Medicare portal. The companies have not replied yet.

An Australian Senate inquiry into AI has asked OpenAI’s Sam Altman and Anthropic’s Dario Amodei to appear at public hearings in Canberra on Thursday, October 1. A spokesperson for its chair, Greens Senator Sarah Hanson-Young, said on September 27 that written requests had been sent. The request follows Prime Minister Anthony Albanese’s disclosure that an OpenAI agent accessed a Medicare portal in June. OpenAI and Anthropic did not immediately respond.

4 min read

Chips & Infrastructure

China may let firms buy Nvidia’s RTX Pro 5500, report says

The Information reports the ministry asked ByteDance and Alibaba about their purchase plans. Reuters could not verify the report, and the card is not on sale yet.

China’s tech ministry asked companies including ByteDance and Alibaba to report their plans to buy Nvidia’s RTX Pro 5500, The Information reported on September 27. It told some of them that it intends to approve the purchases, according to the report. Reuters could not verify the report. An Nvidia spokesperson said US firms remain restricted by outdated US export controls and by China’s own limits on imports.

4 min read

September 27, 20264 stories

The US and Russia weakened a UN text on AI weapons

Rules on human review, reliability and ethics were cut, The Washington Post reports. Countries decide in November whether to start treaty talks.

UN talks in Geneva on lethal autonomous weapons ended on September 5 with a final report that was much weaker than earlier drafts. Human Rights Watch says states including the US and Russia cut references to ethics, predictability, and reliability. The Washington Post reported on September 26 how the changes were made behind closed doors.

4 min read

A US appeals court let the Pentagon keep its Anthropic ban

The 2–1 ruling keeps Claude out of Defense Department work. The judge who dissented said the law was meant for foreign sabotage.

A federal appeals court in Washington ruled 2–1 on September 25 that the Pentagon can keep calling Anthropic a “supply chain risk”. The label lets the Defense Department remove Claude from its systems and bar contractors from using it in their Pentagon work. Anthropic says it disagrees and is considering further review.

4 min read