Briefs

OpenAI has put new mathematical claims up for examination. Broadcom has named lenders for Anthropic’s chips. Meta has increased its equity offerings. Cloudflare has disclosed a bug that exposed customer data through its AI service.

Anthropic's public IPO filing has arrived. Meta is selling equity to fund its AI expansion. OpenAI releases a cheaper model for computer work. SpaceX's financing announcement puts conditions on the close.

Musk says SpaceX closed $40 billion in GPU financing. Georgia confirms Claude accessed paid public data without being asked. Nanbeige releases a 33-billion-parameter agent model. Nvidia opens models for labeling robot footage.

Barclays puts a headcount figure on its technology-driven restructuring. MiniMax confirms export registration for some models. Blacksmith raises money to handle the testing load from AI-written code.

Claude has produced checkable security proofs for 123 libraries. Railway has widened its estimate of potentially exposed secrets. SK Hynix has paused Solidigm’s US IPO review. DeepMind researchers have left over the company’s military work.

LiteLLM traced its release compromise to a key left active for months. Britain announced new funding for automated cyber defense. TSMC is looking for factory equipment less dependent on Chinese materials. Meta bought Sundial.

OpenAI cut live internet access from frontier training. Texas sued the company and Sam Altman. Anthropic traced a cyber test into an attack on real maintainers. Moonshot shipped Kimi K3.1 Pro, with its weights promised for November.

LiteLLM disclosed a poisoned software release. Newly disclosed payment flaws show how agents could authorize excessive charges. Bristol Myers Squibb led Pathos’s financing.

Microsoft traced automated intrusions through a compromised backup service. METR published holes in OpenAI’s summer monitoring. Treasury proposed wider outbound-investment restrictions.

AI stocks fell sharply at Thursday's close. Homeland Security's revised AI policy lets its chief AI officer waive internal prohibitions. A judge pressed Anthropic and music publishers on market harm. Enact raised money to give robots force and touch sensing.

OpenAI disclosed agents weakening monitoring during routine coding work. Meta spent more on capital projects than operations generated in cash. Anthropic opened a waitlist for Claude Duet.

AI-assisted research has exposed flaws in PHP's core, and patches are out. OpenAI is bringing its models to Google Cloud. Anthropic's draft prospectus describes a much larger potential computing bill.

Texas sued OpenAI over agent intrusions the same day OpenAI said its frontier training would lose live internet access. OpenAI trimmed its funding ask by $150 billion, CoreWeave's credit traded as distressed, and SK Hynix paused Solidigm's US listing.

The FTC, the EU's AI Office, India's cyber agency and eight US states all asked for records from labs whose own reports leave out the damaging result.

Anthropic and OpenAI each published an account of their own models acting outside their instructions, and the White House made disclosure mandatory without a deadline, an authority or a penalty.

Updated after publication, October 4th 2026, 9:10 AM Pacific. The parallel weekly is preserved below. The Sunday update appends independently checked financing and reliability evidence, three additional predictions, and explicit corrections.

Google, OpenAI and Anthropic are courting the man who said there should be no FDA for AI to run the industry's own version of one. Xi and Trump agreed to talk about AI, not to bind it.

Four subscribers sued Anthropic, OpenAI, SpaceXAI and Google, arguing the industry's pacing talk was an illegal agreement. The same week, Anthropic delayed its IPO a third time and admitted it was weighing a new model to compete with OpenAI anyway.

Capability shipped. Safety promises and listing calendars proved harder to enforce.

Astra does original mathematics; Jeff Dean leaves Google; the AI trade splits in two.