A powerful new AI model was released this week by U.S.-based Anthropic, maker of the popular Claude chatbot. The new release, titled Claude Fable 5, has novel rules in place to protect against bad actors using the tool for criminal purposes in cyber technology and biotech, including businesses.

Anthropic says Fable 5 uses separate classifiers to detect certain requests tied to cybersecurity, biology and chemistry, or distillation, and routes those responses through Claude Opus 4.8 instead. That safeguard has a tradeoff: Microsoft reportedly restricted employee use of Fable 5 while lawyers review Anthropic’s 30-day data-retention requirement.

Quick housekeeping before the news. We’ve spent the past few weeks rebuilding how this report gets made — the tools, the archive, and where it lives. Soon that will be vorasync.ai. You won’t see any of that work in this issue, and that’s by design.

The report itself isn’t changing — same schedule, same writing. Email delivery is also in the works. We’ll announce the details in a regular issue once everything is ready.

Police in England and Wales were told to stop using commercial AI tools to help draft court-facing documents such as witness statements. (Financial Times)

Software development is surging among individuals with no technical background, data show. (Business Insider)

British officials are warning about deepfakes targeted at government institutions like central banks. (The Guardian)

Attorneys on both sides of a federal lawsuit in Mississippi were disqualified and fined for filings containing hallucinated support. (Mississippi Today; ruling)