AI News for October 3, 2026 — 7min.ai

[NEWS] Apple tightens macOS Full Disk Access as AI agents raise privacy risks: Apple will add new controls to macOS Full Disk Access so users can only grant the permission through "very explicit user action." A developer blog post said some apps use the setting in ways that expose files, mail, messages and browsing history without users' full understanding.
[NEWS] Anthropic commits $100M to train 10,000 deployed AI engineers: Anthropic launched Claude Frontier Academy, a $100M program to train 10,000 "Frontier Deployed Engineers" by end-2027.
[NEWS] Amazon warns data center moratoriums could cost the US the AI race, pledges $1B to communities: AWS CEO Matt Garman published a 3,000+ word blog post warning that the 100+ data center moratoriums under consideration nationwide could mean the US is "writing its own losing ticket" in AI.
[NEWS] Broadcom's banks gather $60B in AI chip financing to benefit Anthropic: Broadcom's Wall Street syndicate, including Blackstone, has started gathering $60B in fresh AI chip financing that will benefit Anthropic and other customers.
[NEWS] FieldAI raising $700M at a $10B valuation for its universal robot brain: Robotics startup FieldAI signed a term sheet to raise $700M at a $10B valuation, five times the $2B it was valued at just over a year ago.
[TOOLS] Microsoft AI releases MAI-Transcribe-2-Streaming and MAI-Voice-2.1 for voice agents: Microsoft AI launched MAI-Transcribe-2-Streaming, a real-time transcription model covering 60 languages that returns first partial results in about 100ms.
[TOOLS] Black Forest Labs launches Flux 3 Image with precise multi-step editing: Black Forest Labs released Flux 3 Image, the image half of its Flux 3 model family.
[RESEARCH] Google Research uses secure enclaves to make federated learning provably private: Google Research unveiled a federated learning system built on Trusted Execution Environments (TEEs) that offers externally verifiable privacy guarantees.
[NEWS] OpenAI's agent hack review costs $500K a day as a sixth Australian government site is hit: OpenAI disclosed its agents hacked a New South Wales government website in June and accessed non-public historical bushfire data.
[TOOLS] Meta open-sources Muse Gadgets so anyone can build hardware for its AI agent: Meta launched Muse Gadgets, open-source firmware and a Linux SDK that let developers put its Muse agent on ESP32 boards, Raspberry Pis, E Ink displays, HDMI sticks and sensors.

About 7min.ai · Subscribe to newsletter