Daily Digest · Archive

AI News for 2026-05-24

The most important AI developments, summarized by AI.

The Decoder

Anthropic may keep supplying Claude to the NSA despite being flagged as a supply chain risk by the Pentagon

Anthropic is likely to continue supplying its Claude AI models to the NSA, even after being flagged as a supply chain risk by the Pentagon. This decision is driven by a shortage of the latest Nvidia Grace Blackwell chips, with Anthropic's "Mythos" model reportedly functioning on older hardware. Crucially, the contentious "any lawful use" clause that previously stalled negotiations will not be part of the agreement.

Key Takeaways

  • Anthropic's Claude AI models may still be supplied to the NSA despite being identified as a supply chain risk.
  • The NSA's reliance on older hardware and a lack of newer chips is a key factor in this decision.
Why it matters: The NSA's continued access to Anthropic's AI, despite supply chain concerns, highlights the critical need for AI capabilities within intelligence agencies and the current hardware limitations they face.
Read Original →
The Decoder

Researchers let Claude Code discover AI scaling algorithms that humans probably wouldn't have designed

Researchers utilized an AI coding agent, AutoTTS, to independently discover novel AI scaling algorithms for reasoning. The discovered algorithm significantly reduced compute costs by approximately 70% while maintaining accuracy, a result likely unforeseen by human designers. This breakthrough was achieved at a remarkably low cost of $40 and a duration of 160 minutes.

Key Takeaways

  • AI agents can discover novel and highly efficient AI algorithms for reasoning.
  • A new algorithm was found that drastically cuts compute costs for AI reasoning while preserving accuracy.
Why it matters: This development suggests that AI agents can automate the discovery of more efficient AI architectures and algorithms, leading to significant cost savings and potentially accelerating AI research.
Read Original →