Overview
- Anthropic CEO Dario Amodei publicly called for slowing frontier model development and for permanent third‑party safety assessments, a stance that drew public backing from major AI figures including OpenAI’s Sam Altman, Elon Musk and DeepMind leaders.
- Microsoft published a detailed 'Humanist AI' draft policy that frames stronger, continuous safety testing as necessary before new capabilities are widely deployed and signals growing industry support for formal safety frameworks.
- Commercial moves are intensifying: OpenAI has held private financing talks targeting a roughly $1.2 trillion valuation, ByteDance reported higher AI spending that cut first‑half profit and took a large bank loan, Broadcom reaffirmed aggressive AI‑chip revenue targets, and Huawei said it has deployed more than 1,000 Ascend supernodes.
- Work on embodied AI continues but faces a data gap: Leapmotor says a self‑developed humanoid is in testing but not yet a commercial focus, and startups such as Lingling Intelligent are raising early funding to scale costly real‑world interaction data collection and governance.
- The mix of safety proposals and heavy capital and compute commitments could raise the cost of compliance and infrastructure, concentrating power with firms and countries that already control chips, large funding rounds and data supply, while some governments urge continued pace to protect local industry.