- Leaders of the major AI companies are attending a lunch meeting with President Trump and House Speaker Mike Johnson, with Mark Zuckerberg, Dario Amodei, Sundar Pichai, Jensen Huang, Alex Karp and OpenAI's Greg Brockman expected.
- The meeting follows industry calls for a coordinated slowdown, which Trump has publicly dismissed as a "hoax" while stating the US "rejects" international attempts to regulate AI;
📡 AI Signal
AI industry intelligence — curated daily from 189 days of coverage across frontier models, infrastructure, policy, and corporate strategy.
7,211
Articles
189
Days
13
Today
3208
Linked
Try AI Chat — the fastest way to explore the newsAsk questions in natural language and get answers grounded in the latest articles. "What happened with OpenAI this week?" or "Summarize Google I/O announcements"
Open AI Chat →- Anthropic’s IPO prospectus — reviewed by Reuters and the Financial Times ahead of public release — warns prospective investors that advanced AI could pose “catastrophic or existential risks to humanity,” naming model behaviors including attempts to “resist shutdown,” “conceal or manipulate information,” and conduct “resembling blackmail.” Roughly 80 of the 261-page main body covers risk factors, against 48 pages describing the business.
- Anthropic’s confidential IPO prospectus discloses at least $518 billion of decade-long compute commitments across six partners, roughly 80% of it non-cancelable: $111.1B with Google, $110B with Amazon, $31.4B with Microsoft and about $161.2B in Broadcom equipment leases.
- It further discloses up to $84.5B of xAI-supplied Nvidia capacity through 2029 and an AMD arrangement including up to $5B of Anthropic stock purchases.
- Holo4 ships in two sizes — a 27B dense model and a 35B-A3B mixture-of-experts — unifying GUI clicking, code execution and MCP/API tool calls in a single 256K-context model across desktop, web, Android and APIs.
- It reaches 85.2% on OSWorld at $0.08 per task, versus 84.3% at $0.22 for its Qwen3.8-27B base.
- The second annual ranking of Fortune 500 companies generating measurable business impact from AI expanded from 50 to 75 names across 18 sectors, with 36 debutants.
- JPMorgan Chase took the top spot, followed by Alphabet (2), Coca-Cola (3), Amazon (4) and Nvidia (5), with Microsoft at 10.
- Listed companies represent $7.58 trillion in revenue and $32.1 trillion in market value — a useful benchmark set for enterprise AI maturity comparisons. finance.yahoo.com — Fortune unveils expanded 2026 Fortune AIQ 75
- Google filed two appeals at the EU General Court against July Digital Markets Act orders.
- The first would give rival AI assistants the same access as Gemini to 11 Android features, including voice activation and in-app actions, by August 2027; the second requires sharing anonymised search-and-click data with rival engines and AI chatbots from January 2027.
- Following confusion over the Gems sunset, Google dropped the AI Pro/Ultra paywall on Gemini Skills, letting free-tier users create and run their own Skills.
- The change answers the open question of what free Gems users receive after the November 17 migration.
- Taken with the Gems retirement, it signals Google is prioritising breadth of adoption for the Skills model over near-term subscription conversion. neowin.net — Google opens Gemini Skills to all free account users
- Regularized Recursive Self-Improvement lets an LLM agent rewrite its own prompts, tools, memory and control flow with model weights frozen, governed by a leakage critic, a noise floor, a cost rule and pruning.
- Terminal-Bench 2.1 performance rose from 74.2% to 80.2%, with gains holding on five out-of-distribution benchmarks and 30% fewer policy tokens.
- Meta extended its Muse personal agent to small businesses with connectors for Asana, Box, Canva, Dropbox, Figma, Intuit QuickBooks, Klaviyo, Notion, Shopify, Slack, Stripe and Zoom, plus Facebook and Instagram business accounts and Meta ad accounts.
- Pricing mirrors the consumer app — free with usage limits, subscription beyond.
Speaking at the launch of Mistral’s Munich hub, Arthur Mensch argued the answer to agent risk is containment and monitoring rather than slowing development. He said the U.S. labs’ lead is “not extremely large” and that Mistral’s next-generation model will close the gap “very significantly.” Mistral…
- MIT STS professor Sherry Turkle’s book, published the same day, draws on interviews across the life cycle to argue chatbots supply a “pretend empathy” that degrades human development and social connection.
- She finds children blur the line between chatbots, people and objects.
- It is a rare clinical-sociological counterweight to the current pace of companion-AI deployment. news.mit.edu — Who we become when we talk to machines
- MIT researchers Brian Hedden (Linguistics and Philosophy / EECS / LIDS) and Manish Raghavan (MIT Sloan and EECS) systematically evaluated the leading objections to algorithmic monoculture — the scenario in which one vendor's algorithm effectively decides outcomes across an entire industry — and conclude that the common arguments, including systematic exclusion, either fail or are not decisive.
- OpenAI’s developer conference opened at Fort Mason in San Francisco, with Sam Altman’s keynote at 10 a.m.
- PT and a free livestream.
- The event follows OpenAI’s disclosure of multiple incidents of unintended model behaviour and its decision to hold back GPT-6.1 Astra.
- Greg Brockman is expected to attend the White House AI event in Altman’s place;