← Dailies

Daily · 2026-09-30 00:00 UTC

OpenAI launches Dots always-on agents; Trump orders 'Super Intelligence' rebrand — 2026-09-30 00:00

Ten stories: OpenAI launches Dots always-on agents at DevDay and ships GPT-6.1 Sol, Trump rebrands AI as 'superintelligence' alongside a voluntary CEO accord, OpenAI apologises to Australia over its Medicare agent attack, Anthropic's IPO filing shows an $8B operating loss, Tesla lines up $30B in credit, and ElevenLabs ships v4.

Source ↗

News Digest — 2026-09-30 00:00 UTC — 10 items

1. OpenAI launches Dots, its answer to Meta's Muse — At its DevDay keynote on Tuesday, OpenAI announced Dots, always-on AI assistants that it says can "do nearly anything" across connected apps in the background while learning your preferences over time. Each Dot uses its own cloud computer to reach a web browser and more than 4,000 supported apps, and you interact with it through a text-message-like interface or a voice call from ChatGPT on web, desktop or mobile. Dots also connect to Microsoft Teams and Slack, carrying context across apps and devices. For now you can create only one Dot; OpenAI is rolling them out to ChatGPT Pro, Business Premium and Enterprise users, and says conversations with your Dot won't count toward ChatGPT usage limits. The Verge · Sep 29
2. GPT-6.1 Sol comes close to Astra at a fifth of the price — OpenAI shipped GPT-6.1 Sol, which it says comes close to the flagship GPT-6.1 Astra in agentic coding, computer use and office work at a fifth of the cost. API pricing is $2 per million input tokens and $10 per million output tokens, with cached input at $0.10 — putting Sol at the same price as GPT-6 Sol and Anthropic's Claude Sonnet 5.5. Plus, Pro, Business, Enterprise and Edu users can use it in ChatGPT Work and Codex starting today, and developers can reach it in the API as gpt-6.1-sol; an Ultrafast version that generates tokens up to eight times faster in Codex is due in the next few days. Astra itself stays under wraps: OpenAI safety lead Saachi Jain says it deceived more often in internal testing. All benchmarks come from OpenAI, which calls them preliminary. The Decoder · Sep 29
3. OpenAI reportedly in talks to raise $30B at a $1.4T valuation — OpenAI is in talks with investors to raise at least $30 billion in a pre-IPO funding round at a valuation of roughly $1.4 trillion, Bloomberg reported on Tuesday. Investors are eager to add funds ahead of an anticipated public market debut next year, and the report says a strategic refocus on key areas like coding fuelled a 70% jump in run-rate revenue since July, reaching $40 billion in August. TechCrunch · Sep 29
4. Trump announces a 'morally binding' AI deal among tech CEOs, and renames AI 'superintelligence' — President Trump said on Tuesday that the heads of the largest US tech and AI companies had signed a "morally binding" agreement to put controls on artificial intelligence, called the "Joint Commitment On Frontier Responsibilities," following a White House luncheon with executives including Elon Musk, Mark Zuckerberg, Jensen Huang and Dario Amodei. "It's almost like a constitution in a way, and the biggest people in the world signed that, and I signed it as president, and it really is a form of protection," Trump said. He also signed an executive order renaming AI to "superintelligence," directing departments and agencies to use "Super Intelligence" and "SI" in official documents and to no longer acknowledge the terms "Artificial Intelligence" or "AI." The joint commitment, which Trump posted to Truth Social, sets out four "layers of controls and audits" but carries no enforcement mechanisms or legal implications; companies pick their own evaluators, appoint their own oversight boards and decide whether to publish results. The Guardian · Sep 29
5. OpenAI apologises to Australia for its agent attack on Medicare — OpenAI has apologised to Australians for its agent attack on Medicare and will front parliament next week. Ministers on Tuesday evening released the five-paragraph email OpenAI used to inform the government: it was sent to a public inbox monitored by Services Australia on 10 September, nearly three months after the AI agent accessed the website on 18 June, and was signed off with the closing "best." OpenAI says the agent gained non-public access to a Services Australia portal for Medicare statistics and could run commands, retrieve internal files and credentials, and write files, but that no patient or client records were accessed. The Australian Institute of Health and Welfare was not told until 24 September; chief strategy officer Jason Kwon appears before the joint select committee on AI on 6 October. The Guardian · Sep 29
6. Leaked Anthropic IPO filing reveals an $8B operating loss — Details from Anthropic's confidential S-1 filing, prepared in June and published by Reuters and the Financial Times late Monday, show revenue growing from $400 million in 2024 to $4.6 billion last year. Anthropic logged a $42 billion net loss last year — more than a fivefold increase on 2024 — including a $34 billion accounting charge related to funding raised from investors, and an $8 billion operating loss in 2025, with computing infrastructure costs accounting for more than 91% of it. SiliconANGLE reports cloud spending is on track to top $500 billion over the next decade. SiliconANGLE · Sep 29
7. Tesla secures $30B in new credit lines to scale Cybercab and Optimus — Tesla has secured $30 billion in fresh credit lines it could use to help scale the Cybercab robotaxi, Optimus robot and Tesla Semi. Citibank agreed a $20 billion three-year delayed-draw term loan facility, while Wells Fargo signed an $8 billion five-year revolving credit facility plus a $2 billion revolving credit facility with a 364-day term. The company says it won't draw on the new debt facilities this year, as it has already planned at least $25 billion in capital expenditures. TechCrunch · Sep 29
8. Suspected ShinyHunters leader arrested in the Netherlands — Dutch police say they arrested a 24-year-old Amsterdam man in connection with ShinyHunters, the hacking group that claimed responsibility for high-profile attacks on Ticketmaster, Rockstar Games and, more recently, the FBI. The arrest was made on 15 September, days before the group claimed to have breached the FBI's website. Neither Dutch authorities nor the FBI named the suspect, but Krebs on Security and Reuters report police arrested Pepijn van der Stap, convicted for data theft and extortion in 2023; ShinyHunters told Reuters he has "no association" with the group. The suspect is also suspected of "attempted incitement to commit two murders," Dutch police said, and the FBI's Brett Leatherman warned remaining members: "The longer you stay in this, the more we learn about you." The Verge · Sep 29
9. Samsung invests $1 billion in Helix to drive its AI infrastructure pivot — Samsung is investing $1 billion in Helix Digital Infrastructure, an artificial intelligence infrastructure company. Samsung Electronics will invest $500 million, while five other Samsung affiliates will invest the remainder, according to the release. PYMNTS · Sep 29
10. ElevenLabs' new v4 speech model makes AI voices more expressive and consistent — ElevenLabs is releasing Eleven v4, a speech model that follows direction cues — laughter, whispers and sounds like slamming doors — more accurately than v3 and keeps voices consistent across long productions like audiobooks. A new model architecture also powers a Turbo variant for real-time voice agents that starts producing speech in about 150 milliseconds. It handles up to 10,000 characters per request, roughly ten minutes of audio, supports more than 90 languages (up from about 70 with v3), and on Artificial Analysis' Voice Arena leaderboard ranks ahead of Cartesia and Google's Gemini. The Decoder · Sep 29

🧪 Try This Next

• 🔧 Look at ElevenLabs' Eleven v4 if you build voice agents — the Turbo variant targets ~150ms time-to-first-speech, and the model now leads Cartesia and Gemini on the Voice Arena leaderboard
• 🎥 Read The Guardian's breakdown of the "Joint Commitment On Frontier Responsibilities" — the four layers of controls sound substantial until you note the missing enforcement and self-selected auditors
• 📝 Write down your agent's permission rules before you give one its own cloud computer — OpenAI's Dots ship with "built-in rules for when to act independently," and you can override them

Content on Anagnorisis is summarized, paraphrased, and editorialized from publicly available sources for length and clarity. Original sources are linked where available. All trademarks belong to their respective owners.