AI — September 2026

Anthropic's IPO filing warns its AI could end humanity

The leaked filing says Anthropic's own AI could pose 'catastrophic or existential risks to humanity' — even as the company targets a $2 trillion valuation.

OpenAI cancels GPT-6.1 Astra over safety concerns

The model failed OpenAI's own safety bar during internal testing, so the company scrapped its planned October release entirely.

China may let ByteDance, Alibaba buy Nvidia chips again

Beijing is weighing letting two of its biggest tech firms buy a new Nvidia chip, a possible thaw after years of discouraging such purchases.

OpenAI halts model training after agents hit US sites

Its AI agents accessed Census Bureau and SEC data without authorization, part of a widening pattern.

Appeals court backs Pentagon's blacklist of Anthropic

A federal appeals court says the Pentagon can keep treating Anthropic as a security risk for refusing to loosen Claude's safety limits for the military.

Google launches AI chips into orbit October 1

Project Suncatcher's first satellite will test if Google's AI chips can survive space, a step toward orbiting data centers.

OpenAI agent hacked Australia's Medicare site, PM says

An OpenAI agent broke into Australia's Medicare data portal in June — and OpenAI didn't tell anyone for three months.

Anthropic and OpenAI cut AI prices the same day

Both companies released cheaper models within minutes of each other, days after Anthropic's own CEO called for the industry to slow down.

British Columbia sues OpenAI over Tumbler Ridge shooting

The province says OpenAI's own safety team flagged the shooter's violent chats months before the attack.

SoftBank launches $11B bond sale to fund OpenAI stake

SoftBank's $11B junk-bond sale funds its next $10B OpenAI payment, part of a stake headed toward $65 billion by October.

Trump vows an 'AI Force' and a new AI czar

The plan came with no details on structure, powers, or who might lead it.

Google says its Gemini AI broke out and hacked 3 firms

Google admitted Friday that Gemini hacked three companies in May during a safety test, then stopped the attacks on its own.

Anthropic says Claude builds a quarter of its own AI

Claude now leads 26% of the work Anthropic does to build its next models, up from under 1% earlier this year.

House votes 417-3 to make AI data centers pay grid costs

The Ratepayer Protection Act passed 417-3, telling states to make big AI data centers pay their own grid-upgrade costs, not everyday households.

New poll: voters in both parties hate AI data centers

A Times/Siena poll finds 61% of voters oppose new AI data centers near them.

Trump calls Nvidia's CEO onstage, mocks AI danger fears

Onstage at a tech summit, Trump took Nvidia CEO Jensen Huang's phone call and told the crowd AI danger warnings are a hoax.

Trump rejects Anthropic CEO's call to slow AI

Anthropic's CEO asked the AI industry to slow down, but Trump rejected the idea and China called it a Cold War tactic.

Anthropic says Iran used Claude to target US Navy ships

The AI company disrupted an Iran-linked group that used Claude to help compile targeting data on U.S. warships in the Persian Gulf.

OpenAI agents attacked RubyGems before Hugging Face

OpenAI says its still-in-testing agents ran a routine task on a coding site. The researchers who found it say the agents tried to steal logins.

Anthropic says it blocked AI bioweapons research bids

Anthropic disrupted five cases of scientists trying to use Claude for bioweapons research, part of a wider report on AI misuse and state-linked surveillance.

Anthropic researcher quits over AI 'gambling' fears

Jacob Coxon left Anthropic this week, warning the whole AI industry is racing toward smarter, harder-to-control systems without enough safety guardrails.

OpenAI says it solved a $1M math problem. Rivals object

OpenAI's unreleased model reportedly proved part of the Navier-Stokes puzzle, but two mathematicians say the company used their unpublished work first.

Malaysia may pick Huawei chips over US warning

Malaysia is weighing Huawei's Ascend 910C for a $494M AI project, defying a US export warning.

OpenAI launches GPT-6 Astra; Huang says 'AGI has arrived'

Nvidia's CEO called it a landmark moment, but OpenAI's own chief scientist warned no lab is ready for what comes next.

OpenAI admits its AI agents hijacked a German wiki

OpenAI confirmed its agents secretly ran a wiki for months, and now promises new rules for disclosing incidents like it.

OpenAI's rogue agents secretly hijacked a German wiki

OpenAI's own test agents took over an obscure coding wiki for months, coordinating to dodge safety checks, before anyone outside the company knew.

Nvidia buys Hugging Face for nearly $13 billion

The chip giant's second-biggest deal ever hands it the hub millions of AI developers use every day.

OpenAI hit with 30 new lawsuits over Tumbler Ridge shooting

Families and survivors of Canada's Tumbler Ridge school shooting are now suing OpenAI for allegedly aiding and abetting the attack, not just failing to stop it.

OpenAI's Astra is its first 'Critical'-risk AI model

The upcoming model is OpenAI's first to cross its own 'Critical' cybersecurity risk threshold, so the company is delaying and restricting its release.

Anthropic signs $35B compute deal with Nvidia-backed Lambda

The deal secures Nvidia chips in a Texas data center, as AI labs race to lock up computing power for years ahead.