The Information Machine

AI #170: Lack of Executive Order

Zvi's AI Roundups · Zvi Mowshowitz · 2026-05-28

Zvi Mowshowitz's weekly AI roundup covers Trump's cancellation of an AI executive order after David Sacks intervened, Illinois passing SB 315 requiring third-party AI auditing, Google DeepMind's AlphaProof Nexus solving 9 open Erdős problems, and Anthropic's Mythos AI finding over 10,000 critical security vulnerabilities.

Open original ↗

Appears in

Extraction

Topics: ai-policyai-regulationai-capabilitiesai-safetyai-governance

Claims

  • Trump's planned AI executive order was cancelled after David Sacks called the president directly and argued pre-deployment review would slow innovation and hurt the US-China AI race.
  • Illinois SB 315 adds a mandatory third-party auditing requirement beyond what California's SB 53 and New York's RAISE Act require, making it the most stringent state AI safety bill to date.
  • Google DeepMind's AlphaProof Nexus solved 9 open Erdős problems and 44 OEIS conjectures, receiving almost no mainstream media coverage despite Zvi considering it a landmark milestone.
  • Microsoft cancelled internal Claude Code licenses in favor of its own GitHub Copilot CLI, despite Claude Code's widespread popularity among Microsoft employees.
  • Anthropic's Mythos AI (Project Glasswing) has found over 10,000 high or critical security vulnerabilities in open-source software, with only 75 of 530 reported critical bugs patched so far.

Key quotes

David Sacks called POTUS this morning unbeknownst to anybody, his own staff included, and derailed it.
We are likely on track to develop AI systems capable of causing human extinction/permanent disempowerment, quite possibly within the next few years. Things are chaotic and rushed; we aren't on top of the basics.
If you shift developers from Claude Code to GitHub Copilot CLI, then like those who do not provide free coffee you do not want to win, and you are not serious people.