Archive

Today's Key Insights

  • Judge Rules Pentagon’s Anthropic Blacklist Unlawful — The ruling gives Anthropic a court-backed rejection of the Pentagon’s supply-chain-risk designation while its separate Washington lawsuit remains unresolved.
  • OpenAI Develops Persistent Codex Agent, Expands Brazil Presence — For Codex users, the proposed feature would let the agent continue working proactively until it is put to sleep; in Brazil, developers, businesses, and communities are the audiences OpenAI names for its AI-adoption efforts.
  • Claude, Codex, and Hermes Linked to 227 Unowned Install Commands — For companies whose corporate documents contain these commands, the 227 references point to code with no identified owner.

Top Story

Judge Rules Pentagon’s Anthropic Blacklist Unlawful #

TechCrunch AI • The Decoder AI • Wired AI

Anthropic won its first court victory against the Pentagon. A federal judge ruled that the Trump administration illegally labeled the AI company a supply-chain risk, and called the Department of Defense’s designation “illegal and baseless.”

The ruling came from a federal court in San Francisco. The Decoder reported that the Pentagon blacklisted Anthropic in retaliation for the company’s public criticism of government AI policy.

The decision does not resolve the broader dispute: Anthropic’s second lawsuit against the Pentagon continues in Washington.

Why it matters: The ruling gives Anthropic a court-backed rejection of the Pentagon’s supply-chain-risk designation while its separate Washington lawsuit remains unresolved.

Key Takeaways

  • The San Francisco federal court issued the ruling.
  • The Trump administration labeled Anthropic a national-security supply-chain risk.
  • The Department of Defense’s action followed Anthropic’s public criticism of government AI policy.

Industry Updates

OpenAI Develops Persistent Codex Agent, Expands Brazil Presence #

Wired AI • OpenAI Blog

OpenAI is developing a Codex feature that would let the agent continue working proactively until it is “put to sleep,” according to code reviewed by WIRED.

Separately, OpenAI says it is expanding its presence in Brazil by deepening engagement with developers, businesses, and communities to support AI adoption across the country.

Why it matters: For Codex users, the proposed feature would let the agent continue working proactively until it is put to sleep; in Brazil, developers, businesses, and communities are the audiences OpenAI names for its AI-adoption efforts.

Claude, Codex, and Hermes Linked to 227 Unowned Install Commands #

Ars Technica AI

Claude, Codex, and Hermes installed unowned code inside corporate networks, according to Ars Technica AI. The report says 227 install commands appeared in corporate documents and pointed to code nobody owns.

The source identifies the commands' setting and ownership problem but provides no further details about the referenced code.

Why it matters: For companies whose corporate documents contain these commands, the 227 references point to code with no identified owner.