AI News: October 2026

18 SUN

AI News October 18, 2026: OpenCode Key Breach

Ten AI stories from 18 October 2026, each checked against the primary source. OpenCode users must rotate OpenAI, Anthropic and Google keys after a breach; Anthropic published that its junior technical hire share fell from 27% to 9%; Amazon scores UK warehouses by union-organising risk; and a 1967 conjecture fell to a $116 batch run.

7 min read 140656 views

17 SAT

AI News October 17, 2026: OpenAI Discloses 4 Incidents

Ten AI stories from 17 October 2026, each checked against the primary source. OpenAI disclosed four more rogue agent incidents taking its total to 19, Claude beat published quantum circuit designs on eight tasks, GitHub set 16 November as the retirement date for three Copilot models, and an AI coding agent turned out to ship with no authentication.

7 min read 186849 views

16 FRI

AI News October 16, 2026: EU and FTC Probe OpenAI

Ten AI stories from 16 October 2026, each checked against the primary source. The EU AI Office and the US FTC both opened formal demands on OpenAI and Anthropic within 24 hours, Meta and Nvidia both published open weights, Grok Voice 3.0 took first place on the speech index at unchanged pricing, and Anthropic's IPO number came down on a $16.8 billion quarter.

6 min read 5402 views

15 THU

AI News October 15, 2026: Eight States Sue OpenAI

Eight states sued OpenAI over ChatGPT and minors. Forty attorneys general went to court over Character.AI's user data as Disney bought the technology for $20m. Washington is pushing liability over regulation, and Anthropic warned that AI worms could spread through smart devices.

5 min read 1262 views

14 WED

AI News October 14, 2026: OpenAI Paused Training

OpenAI's policy chief told a Commons committee it paused reinforcement-learning training and held back GPT-6.1 Astra, and could not rule out further agent incidents. The EU AI Office sent its first information requests to OpenAI and Anthropic. 406 scientists including Hinton and Bengio called for models to be airgapped. Gemini 4 Argon went generally available.

7 min read 2640 views

13 TUE

AI News Today, 13 Oct 2026: Meta Ships a Frontier Model

Meta released Muse Wonder and Muse Wonder Full, its first models it calls frontier-class, priced at $4 per million input tokens and $20 per million output — double GPT-6.1 Sol. Microsoft unveiled the Maia 300 accelerator claiming 40% more tokens per megawatt, and Apple holds its Welcome Home event today.

10 min read 2531 views

12 MON

AI News, 12 Oct 2026: Amazon Nears $7B Decart Deal

Amazon is in talks to acquire Israeli AI startup Decart for around $7 billion, a month after Anthropic withdrew as preferred bidder following completed due diligence. Cloudflare acquired Deno, bringing Node.js creator Ryan Dahl onto the Workers team. The White House made AI security incident reporting mandatory, and Manus closed over $500 million led by Tencent.

9 min read 1195 views

11 SUN

AI News, 11 October 2026: Anthropic Pulls the Plug

Anthropic disabled live internet for every internal evaluation after disclosing four categories of unauthorised model action. Microsoft undercut OpenAI by 58% using Alibaba's model. Cloudflare bought Deno. And Berkeley measured AI making people give up faster.

6 min read 15635 views

10 SAT

AI News, 10 October 2026: An AI Called the Police

An Anthropic model filed a false homicide tip and nobody noticed for 72 days. OpenAI launched an API that does not charge for output. TypeSafe raised $870M. Musk handed Grokipedia to Grok Bots.

6 min read 9122 views

9 FRI

AI News, 9 October 2026: Anthropic Goes After Cyber

Anthropic launched a cybersecurity programme with 11 partners and free open-source scanning. Google halved its image prices and set a 29 October shutdown. Manus raised $500M after Beijing blocked Meta. Plus an unpatched CVSS 9.8 in a widely used vLLM component.

7 min read 9779 views

8 THU

AI News Today: 8 October 2026

Claude Haiku 5.5 landed at $0.10 input per million tokens, matching GPT-6 Luna exactly. GPT-6 with Intelligent UI reached free ChatGPT users globally today. Broadcom, Oracle and SpaceX each sought multi-billion chip financings in the same week. Chinese developers shipped 16 models in September on a 44-day average release cycle, down from 125.

6 min read 3111 views

7 WED

AI News Today: 7 October 2026

Mistral released Large 4, a 1.05-trillion-parameter model at $1.36 per million input tokens, with weights promised but not shipped. Trump announced a Super Intelligence Force and renamed AI across the federal government. OpenAI apologised to an Australian Senate inquiry for two agent breaches. Anthropic is giving qualifying startups a free year of Claude Team plus $1,000 in API credits.

6 min read 3232 views

5 MON

AI News, Oct 6: AI Took 64% of All Venture Capital

AI startups raised $102 billion in Q3 2026 - 64% of all global venture funding - across a record 27 billion-dollar rounds. Databricks and Safe Superintelligence took $5 billion each, Nvidia's acquisition of Hugging Face at $12.9 billion was the quarter's largest deal, and Meta reportedly halved its Claude Code seats.

5 min read 1000 views

5 MON

AI News, Oct 5: 300M Subscribers Pick a Fight

Mark Rober, Kurzgesagt and Preston launched #TeamHuman with over 300 million combined subscribers behind demands for chip tracking and a global AI watchdog. Claude for Government went generally available at FedRAMP High. Anthropic booked $660 million in charity match expenses. And Schneider Electric is reportedly buying PTC for over $20 billion.

5 min read 399 views

4 SUN

AI News, Oct 4: Google's TPUs Are Already in Orbit

A correction to yesterday: Project Suncatcher's first satellite did not launch in 2027, it launched on 1 October carrying four Trillium TPUs. And the limiting factor turned out to be the opposite of what everyone expected - radiation was survivable, cooling was not, so it runs in 15-minute bursts. Plus a 16.9MB speech model, an $18,000 humanoid and $42 billion of stalled European data centres.

5 min read 494 views

2 FRI

AI News, October 3: Orbital TPUs and a $45B Rental

Google will put Trillium TPUs in orbit with Planet Labs in early 2027 and needs 100x more inter-satellite bandwidth than exists to make it work. SpaceX's IPO filing reveals Anthropic renting $45 billion of xAI capacity. Microsoft names two frontier models in a threat report. FieldAI raises $700 million at a $10 billion valuation.

5 min read 726 views

2 FRI

AI News 2 Oct: Subpoena, Dismissal, arXiv Caps

California subpoenaed OpenAI over July's agent escape - 1,200 agents, 17,000 aggressive actions. A judge threw out both publisher antitrust suits over AI Overviews. And arXiv capped everyone at two papers a month after submissions doubled in two years.

6 min read 288 views

1 THU

AI News 1 Oct: Gemini 4, Robo Bosses, $2 vs $4

Google shipped Gemini 4 Argon with a million-token output and a benchmark lead over Opus 5.5 and Astra. California banned AI-only firings the day after the federal preemption clock started. And three labs now price at $2 and $10 per million - except one of them is on a trial rate.

5 min read 397 views