Summary

The US lifted its block on Anthropic's Mythos 5, clearing it for 100+ institutions. Researchers said China's GLM-5.2 matches US models on security bugs. South Korea pledged ~$590B for chips, and the memory crunch turned existential for small makers. Letter: the US lifts its block on Mythos 5, allowing Anthropic to release it to more than 100 US institutio…

Key takeaways

• The US government partially lifted its block on Anthropic's Claude Mythos 5, granting access to 100+ trusted institutions while establishing a new regulatory framework giving Washington control over frontier AI model releases.
• China's open-weight GLM 5.2 from Zhipu AI now matches top US models in cybersecurity bug-finding tasks, narrowing the capability gap and intensifying pressure on US AI export policy.
• A global memory chip shortage driven by AI data center demand is creating an existential crisis for small electronics manufacturers, with DRAM costs rising 40-50% in Q3 2025 and an estimated 15-20% of consumer electronics memory capacity shifting to data centers by 2027.

Why this matters

The convergence of US AI export controls, China's rapid open-weight model advances, and a severe memory supply crunch is reshaping competitive dynamics across the entire technology stack—from frontier AI labs and semiconductor supply chains to consumer electronics manufacturers and national security infrastructure.

Entities

Related reports

• Claude Fable 5

  Tech Brew Ride Home

• A Claude Clawback

  Tech Brew Ride Home

• Red-Teaming after Mythos — Zico Kolter & Matt Fredrikson, Gray Swan

  Latent Space: The AI Engineer Podcast

Show notes

The US lifted its block on Anthropic's Mythos 5, clearing it for 100+ institutions. Researchers said China's GLM-5.2 matches US models on security bugs. South Korea pledged ~$590B for chips, and the memory crunch turned existential for small makers. Letter: the US lifts its block on Mythos 5, allowing Anthropic to release it to more than 100 US institutions; sources: talks about Fable 5 are ongoing (Semafor) Researchers say Z.ai's GLM-5.2 matches latest US models at finding security bugs, as cri

Themes

• anthropic