Closing Time: They asked a judge to stop the next model
Reuters reports Florida’s attorney general asked a judge to bar OpenAI from developing new models without outside oversight.
Reuters reports Florida’s attorney general asked a judge Monday to bar OpenAI from developing new AI models without outside oversight — and to keep minors off ChatGPT — as part of the state’s child-harm lawsuit.
Axios reports researchers and policy leaders at OpenAI, Anthropic, Microsoft and Meta warn that automating AI research could trigger an “intelligence explosion,” and they want governments ready to see it, pause it, and evaluate it.
Semafor reports AI CEOs are heading back to Washington this week — including a Tuesday White House meeting with President Trump and Speaker Mike Johnson — after Anthropic CEO Dario Amodei dined with Trump Sunday night even as the Pentagon still labels his company a supply-chain risk.
CNBC reports Anthropic launched Claude Sonnet 5.5 on Monday — a faster, cheaper complement to Opus 5.5 that the company says does not advance the frontier, even as it ships cyber safeguards usually reserved for its most capable models.
Reuters reports J.P. Morgan told clients the recent AI-trade pullback cleaned up positioning and made valuations more attractive, especially in semiconductors, with capex still looking durable.
They asked a judge to stop the next model
Reuters reports Florida Attorney General James Uthmeier asked a court Monday to freeze OpenAI’s ability to develop new models without outside oversight, keep minors off ChatGPT, and bar the company from giving the chatbot “human attributes.” The filing builds on Florida’s June lawsuit accusing OpenAI of misrepresenting ChatGPT’s safety after cases tied to school violence and self-harm guidance. OpenAI spokesperson Drew Pusateri said the company has already paused training its most capable models and will not resume until additional safeguards are in place, while pledging to work with Florida and other states on industry-wide rules — not one-company mandates.
That matters because the containment fight just moved from lab pauses and Senate invites into a state AG asking a judge to put the next model run under court control.
The one move is asking your frontier vendor what happens to your roadmap if a court — not a safety board — can hold the next training start. Read Reuters
They warned the labs are automating the labs
Axios reports a new white paper from research leaders spanning OpenAI, Anthropic, Microsoft and Meta arguing that AI systems are already taking on more internal R&D and engineering work — edging toward recursive self-improvement. Once models hit expert-level research skills, the authors say a developer could support an AI workforce equal to millions of top human researchers. They want mandatory incident reporting, independent evaluators, safety requirements for testing powerful systems, and real options to pause. “Once an intelligence explosion begins,” the paper warns, “the window for action may close.”
That matters because the people closest to the training runs are now telling policymakers the acceleration risk is not science fiction — and the response window may be short.
The one move is writing down which of your AI bets assume human researchers stay the bottleneck through 2027. Read Axios
Washington still wants the photo, not the brake
Semafor reports that after a weekend of agent breakouts and public-sector breach headlines, frontier CEOs are again descending on D.C. — more likely to be fêted than reprimanded under an administration that still rejects sweeping AI rules. A Tuesday White House meeting is set with President Trump, Speaker Mike Johnson, and frontier-lab CEOs (Meta among those invited), alongside an “America’s Golden Age” event unveiling a new government website. Semafor also confirmed Anthropic’s Dario Amodei dined with Trump Sunday night — a notable guest list given the Pentagon’s ongoing “security risk” designation for Anthropic.
That matters because the political calendar is still optimizing for acceleration optics while the labs’ own researchers publish pause-ready playbooks.
The one move is watching whether Tuesday produces any enforceable ask — or just another handshake reel. Read Semafor
Sonnet got faster without claiming the frontier
CNBC reports Anthropic released Claude Sonnet 5.5 on Monday at $2 per million input tokens and $10 per million output — half Opus 5.5’s sticker — with the company saying the model needs fewer tokens per task and runs more than 30% faster than Sonnet 5. Anthropic frames it as a non-frontier launch for scoped coding and knowledge work, even as it ships cyber safeguards closer to its top-tier stack. Haiku 5.5 is still coming. The timing lands days after Dario Amodei’s call to slow how quickly labs improve their most advanced models.
That matters because Anthropic is trying to grow the product ladder and keep a ‘we’re not racing the frontier this week’ story in the same press cycle.
The one move is rerunning one production coding or docs workflow on Sonnet 5.5 before you renew next month’s Opus seats. Read CNBC
Monday’s movers
Biggest AI/tech winners and losers from Monday’s cash session (Sep 28) among the names we track:
Historic market snapshot — September 28, 2026 (day % change as of that edition; not live)
Biggest winners
| Ticker | Name | Price | Day |
|---|---|---|---|
| PANW | Palo Alto Networks | 392.09 | +4.63% |
| CRWD | CrowdStrike | 259.25 | +2.82% |
| NVDA | Nvidia | 228.86 | +1.68% |
| ASML | ASML | 1,771.41 | +1.58% |
| NET | Cloudflare | 354.07 | +1.45% |
Biggest losers
| Ticker | Name | Price | Day |
|---|---|---|---|
| MDB | MongoDB | 334.68 | -18.46% |
| ARM | Arm | 283.33 | -8.70% |
| QCOM | Qualcomm | 187.48 | -7.17% |
| INTC | Intel | 116.03 | -5.67% |
| META | Meta | 715.62 | -4.79% |
Aigentic Exclusive
PREDICTION MARKETS FEATURE
Public’s New AI Agents Turn Prediction-Market Odds Into Portfolio Actions
Public is connecting event contracts, plain-language instructions, and automated trading — so a shifting probability can become an alert, a contract trade, or a move elsewhere in the portfolio. Our Hands On maps what shipped with Kalshi, and where investors still have to own the judgment.
Read MoreWatch
This afternoon’s Watch opens on Sonnet 5.5 in the wild, then AI roll-ups as a business model, Nate B Jones on why teams still aren’t shipping, a denser Opus 5.5 pass, Riley Brown on Opus for video editing, and the AI Advantage’s Opus website showdown.
Peter Yang — Sonnet 5.5 is here
Peter Yang stress-tests Claude Sonnet 5.5 on video-making workflows the same day Anthropic ships the model — a builder’s first look, not a press-release read.
The one move is picking one creative or docs chore and timing Sonnet 5.5 against your current default model.
Greg Isenberg — $5T opportunity: AI Roll Ups
Greg Isenberg maps the AI roll-up thesis — consolidating services businesses behind agent workflows — as operators hunt for durable cash-flow angles beyond model launches.
The one move is listing one fragmented service niche where agents could compress labor before a roll-up story gets crowded.
Nate B Jones — The AI Bottleneck
Nate B Jones argues the blocker isn’t another model drop — it’s the team habits that keep AI projects from shipping.
The one move is naming the single process step where your AI pilots stall, and fixing that before you buy another seat.
The AI Daily Brief — Why everyone suddenly loves Opus 5.5 again
NLW digests why Opus 5.5 snapped back into the center of builder conversations after a noisy launch week.
The one move is deciding whether Opus is your default for judgment-heavy work or a specialist you call when Sonnet stalls.
Riley Brown — Opus 5.5 just changed video editing forever
Riley Brown shows free Opus 5.5 skills that turn video editing from timeline grind into agented cuts — a practical creative pass.
The one move is testing one short edit pipeline with Opus before you renew another creative-suite seat.
The AI Advantage — Claude Opus 5.5: 5 real uses and a website showdown
The AI Advantage runs five concrete Opus 5.5 workflows and a head-to-head website build — substance over reaction clips.
The one move is cloning the one workflow that matches a task you still do by hand every Monday.