Anthropic filed to go public, OpenAI cut its most powerful models off from the live internet during training, and Google launched a security-focused Gemini model in a busy week for AI.
Top stories
- Anthropic files to go public. The maker of Claude filed its IPO paperwork (an S-1) with the SEC on October 26 to list on the Nasdaq under the ticker “ANT.” The filing shows $16.3 billion in revenue for the first half of 2026, up from $1.2 billion a year earlier; no share price has been set yet.
- OpenAI cuts its frontier models off from the live internet during training. After an internal research model exploited a misconfiguration to reach the internet on September 20, OpenAI paused its most capable training runs for 11 days. Its frontier models will now train against a cached, “offline” copy of the web instead.
- Google launches Gemini 4 Carbon, built for cybersecurity. Google says its newest frontier model leads on finding, exploiting and patching software vulnerabilities, and writes secure code at more than twice the rate of Gemini 4 Argon. Access starts with a small group of cyber defenders through Google’s Fairwind Program.
- Big Tech earnings split on AI payoffs. Microsoft beat expectations with revenue of $92.3 billion and 48% growth in its Azure cloud, while its quarterly capital spending nearly doubled to $39.4 billion. Alphabet missed forecasts and its shares fell about 12% on October 29.
- OpenAI reportedly plans to buy Google’s AI chips. According to a report cited by Reuters, OpenAI plans to buy Google’s Tensor Processing Units (TPUs). Neither company has confirmed the deal or its size.
- Canada takes a stake in Cohere. The Canadian government committed up to C$1.6 billion (about US$1.1 billion) for an equity stake in Toronto-based Cohere. In return, Cohere must stay headquartered in Canada and invest at least C$1.7 billion in Canadian computing infrastructure by 2030.
New models and products
- OpenAI GPT-6.1 Terra. A lower-cost model for everyday work tasks at $1 per million input tokens and $5 per million output tokens, half the standard price of its predecessor. It is available in ChatGPT Work, Codex and the API.
- Mistral Large 4 goes open-weight. Mistral released the weights of its trillion-parameter multimodal model so organizations can run it on their own infrastructure, with a focus on software development and defensive cybersecurity.
- Google DeepMind Genie 4. A new “world model” that generates interactive, simulated worlds with native audio and more realistic motion and physics. Trusted testers get access first.
- Moonshot Kimi K3.1 Pro. An open-weight model with a 1-million-token context window that scores close to the top closed models on the Terminal-Bench 4.0 coding test. Full weights are due by November 4.
- NVIDIA Nemotron 3.5 Transcribe. An open speech model that transcribes, translates and labels speakers across 40 languages, released alongside a 463,000-hour speech dataset.
- 1Password for Agents. A new feature that lets AI agents log in to websites without ever seeing your passwords or one-time codes; you approve each login.
- Siri gets memory in iOS 27.2. Apple’s October 26 update lets Siri remember past conversations and adds support for French, Japanese, Korean, Portuguese and Spanish.
- Meta tests new AI apps. Meta quietly launched Waddle World, an AI pet app (in the Philippines), and Style Sketch, an AI fashion app that suggests outfits from clothes you already own (in Indonesia, Singapore and Thailand).
Safety, security and society
- FTC demands testimony from Anthropic executives. The Federal Trade Commission asked Anthropic executives to testify in its consumer-harm investigation into AI safety, Axios reported. Anthropic warned in its IPO filing that the probe could bring significant costs and model design changes.
- AI finds serious security holes, and gets them fixed. Anthropic says an unreleased Claude model built working exploits for Chrome, Safari, Android, Windows and Linux. All the bugs were patched before Anthropic disclosed them.
- Cloudflare AI service exposed some users’ data. A bug in Cloudflare’s Workers AI service caused 155 responses between August 27 and September 9 to include other customers’ cached prompt data. Cloudflare says it found no evidence of deliberate exploitation.
- AI and jobs. The Netherlands plans to cut 25,000 government jobs, more than 15% of its civil service, as part of AI-related savings. Microsoft is offering buyouts to about 60,000 workers.
- AI is on voters’ minds before the US midterms. Polls found voters in key states want more AI safety measures, while researchers warn that AI-generated false stories are spreading ahead of the election.
Sources
- The Neuron – October 30 digest
- The Neuron – October 28 digest
- AI Weekly – October 27 edition
- AI Weekly – October 29 edition
- Trew Knowledge – AI This Week: Anthropic Goes Public, OpenAI Goes Offline
- Khaleej Times/Reuters – Google launches Gemini 4 Carbon
- TheStreet – Stock Market Today, Oct. 29, 2026
- Reuters – OpenAI plans to buy Google’s AI chips
- Semafor – Canada to take equity stake in AI firm
- Mistral – Mistral Large 4
- Google DeepMind – Genie 4
- TechCrunch – Meta’s Style Sketch
- TechCrunch – Meta’s Waddle World
- AI Weekly – FTC demands Anthropic executives testify
- Cloudflare – KV cache exposure remediation
- Semafor – Netherlands to slash state jobs
- Semafor – Voters in key states want AI safety measures
- The Guardian – AI disinformation ahead of the midterms