Artificial Intelligence

Turns out measuring AI chart skills is harder than measuring the charts themselves.

During the Sri Lanka cyclone that killed 410 people, ChatGPT usage spiked 17x. Nobody checked whether the answers were right. OpenAI is now building disaster response tools on that unverified foundation.

Five days after a federal judge blocked the Pentagon from blacklisting Anthropic, CEO Dario Amodei flew to Canberra and signed an AI safety MOU with the Australian government — while actively suing the Trump administration.

OpenAI is losing money and making more of it than ever. These facts are not contradictory — they're the business model.

3M downloads a month isn't usage—it's dependency. TRL's breaking changes now cascade instantly into thousands of downstream projects.

Anthropic will 'explore' Australian data center investments. Targeting, not committed. The PM shook hands anyway.

They laughed at it two years ago. Now the same meteorologists are quietly relying on it.

The injunction buys time. But $19 billion in projected annual revenue against $5 billion actually collected is a gap that judges cannot close. Add back-to-back security incidents, and the month is messier than the headline suggests.

A supply chain attack hit a library with 400M monthly downloads, staging three OS-specific payloads 18 hours in advance. This is the same campaign that hit LiteLLM one week earlier. OpenClaw was caught in the blast.

A developer's AI coding assistant updated a library as part of its normal workflow. That update was the attack.

Crosby runs 8 AI agents through every contract before a human lawyer signs off. The agents are fast. The bottleneck is the lawyer.