AI Digest

The US export-control standoff over Anthropic's newest models drags on.

Two weeks after Washington forced Anthropic to pull Fable 5 and the Mythos cybersecurity model over national-security concerns, no one can articulate exactly what rule was broken, and security researchers have signed an open letter calling the ban counterproductive. Reporting suggests the restriction may even be boosting Anthropic's brand, while on Amazon Bedrock the models now require sending prompts and outputs to Anthropic for review.

Sources Wired ·TechCrunch ·InfoQ ·LessWrong

The AI compute race intensifies as rivals chip away at Nvidia.

Amazon is reportedly in talks to sell its in-house AI chips to outside data centers, a move CEO Andy Jassy frames as a $50 billion opportunity, while startups like Tensordyne bet on logarithmic math to undercut GPUs. Capital keeps pouring into the inference layer, with Baseten said to be raising $1.5 billion just months after its last mega-round, even as the US probes whether ASML's top tool reached China.

Sources TechCrunch ·The Register ·TechCrunch

Agentic AI is being baked into mainstream enterprise software.

Adobe is embedding "creative agents" across Premiere, Photoshop and Illustrator, GitLab 19 extends agents into secrets, merge requests and supply-chain scanning, and both AWS Bedrock AgentCore and Azure Functions shipped production agent runtimes this week. The shift is from one-off generation toward agents that orchestrate multi-step workflows inside the tools teams already use.

Sources VentureBeat ·InfoQ ·AWS ·InfoQ

The same agent boom is exposing a wave of security holes.

Researchers found roughly 7,000 Langflow servers under active attack and showed LangGraph and LangChain share the same exploitable bug class, turning ordinary flaws into remote code execution next to sensitive keys. Parallel disclosures hit Microsoft 365 Copilot, via the SearchLeak mailbox-exfiltration chain, and LiteLLM handing out admin keys, with the common thread being AI systems that trust external input without a boundary.

Sources VentureBeat ·VentureBeat ·The Register

Governments move to rein in AI, from classrooms to wealth.

Norway imposed a near-total ban on AI in elementary schools, the UK pressed ahead with face-scanning age checks for asylum-seekers despite knowing the tech is flawed, and Senator Bernie Sanders floated a $7 trillion plan to give Americans a stake in the AI industry. The measures span education, surveillance and economics, signaling a broadening regulatory appetite that the biggest AI firms are likely to resist.

Sources Reuters ·Ars Technica ·Ars Technica

Physical AI consolidates as robotics money flows.

Hyundai took full control of Boston Dynamics, buying out SoftBank's remaining stake for $325 million, while Taiwan ramps drone production for both its own defense and US military demand amid pressure from China. Industry data backs the momentum, with the IFR reporting double-digit growth in US robotics in 2025, led by food and other non-manufacturing sectors.

Sources Startup Fortune ·Ars Technica ·The Robot Report