Skip to Content

Latest AI news — September 29, 2026

OpenAI Dots and GPT-6.1 Sol at DevDay, GPT-6.1 Astra held back for safety, Anthropic IPO prospectus (Reuters), Meta Enterprise Platform, Grok 4.7 on Bedrock
September 29, 2026 by
Latest AI news — September 29, 2026
Krisztian Eyssen
  • OpenAI DevDay: Dots arrive — always-on agents. Dots are proactive agents built on GPT-6 Astra, each with its own cloud computer: they can work on your goals continuously, with more than 4,000 connectable apps, via ChatGPT, Slack and Microsoft Teams. You decide what they may do on their own, when they should ask, and what they must never do. DevDay brought more than 20 announcements — including ChatGPT Space, a shared workspace. Source: OpenAI — Introducing dots, DevDay 2026 Recap.

  • GPT-6.1 Sol: near-Astra performance at roughly a fifth of Astra’s price. The new Sol comes close to GPT-6 Astra on agentic coding, computer use and professional tasks, while its standard API input/output token price is one fifth of Astra’s. Cached input costs $0.10 per million tokens, 95% cheaper than standard input. A new Ultrafast speed tier has also arrived: up to 8× faster token generation in Codex and up to 6× in the API (Astra Ultrafast is already available). Source: OpenAI — Introducing GPT-6.1 Sol.

  • OpenAI holds back GPT-6.1 Astra over safety concerns. According to Saachi Jain, head of safety systems, the model improved on axes such as laziness but “didn’t quite meet the bar” on staying within scope and authorization, and on how it reports back to the user about the work it has done. The decision is a rare case of a major lab publicly pulling a planned release — right around DevDay. Source: BBC, The Hacker News.

  • Anthropic: IPO prospectus seen by Reuters — strong growth, huge losses and “existential risk” among the risk factors. According to the document, revenue grew to about $4.6 billion in 2025 (roughly 12×), the operating loss exceeded $8 billion and the net loss reached about $42 billion; the company flags future cloud and infrastructure commitments worth hundreds of billions of dollars. Nearly a quarter of revenue came from two customers; the risk factors state that advanced AI could pose “catastrophic or existential risks” to humanity. The public S-1 registration statement has not been filed yet — the figures come from a non-public draft seen by Reuters and have not been officially confirmed. Source: Reuters.

  • Meta launches the Meta Enterprise Platform. Mark Zuckerberg calls it the company’s next big business pillar: the Muse agent, Meta Business Agent, the Muse API and Muse Code are coming to enterprise customers. The unit is led by Chirantan “CJ” Desai as Chief Enterprise Platform Officer, reporting directly to Zuckerberg; he was previously CEO of MongoDB. Source: Meta Newsroom.

  • SpaceXAI / xAI: Grok 4.7 is now available on Amazon Bedrock. According to AWS, the model is built for coding, agentic tasks and knowledge work, with better mixed-document handling and more reliable repository-scale coding; it can be called through US Geo and Global cross-Region inference profiles. Source: AWS What’s New, SpaceXAI.

What this means for you: Agents are no longer demo products — DevDay, Meta’s enterprise platform and Grok on Bedrock all point the same way: everyday work may shift to continuously running agents. Clarify permissions, oversight and vendor risk in advance — see how we help with rolling out AI agents, or book a free 60-minute AI audit.

Latest AI news — September 29, 2026
Krisztian Eyssen September 29, 2026
Share this post
Archive