ChatGPT appears to block direct requests to copy an author's style, instead offering to capture the overall "feeling", amid its legal battles with book authors
THE SO WHAT
Style-cloning guardrails in ChatGPT are a live example of model behavior being tuned by litigation risk, not just safety research. If your workflows depend on close imitation of specific voices, plan for more of these soft bans and diversify across tools and in-house models.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIKimi K3's full weights are here, but they're 'open' with a caveat: What enterprises should know
“Open” with a custom license is now a distinct category of model risk—Kimi K3’s full weights are useful, but the attached terms become part of your compliance surface. Before you integrate, have legal treat the K3 license like any other critical vendor contract, not a GitHub README.
Applied AIAnthropic’s Amodei Rejects Open Model Ban But Calls for Testing
Top lab leaders arguing against a blanket open-weight ban while calling for mandatory testing is a pivot toward “regulated openness” as the compromise. If you rely on open models, start documenting evals and risk controls now—those artifacts are likely to be what regulators and partners ask for first.
Why this AI travel company is bringing Sonder back from the dead
Reviving a known consumer brand as an AI-first discovery layer is a bet that distribution and trust matter more than raw model quality in travel. If you’re in a legacy category with a dormant or distressed brand, pairing it with AI-native UX may be a faster route to demand than building a new name from scratch.
Dario Amodei rejects speculations that Anthropic wants to ban all open-weight AI models
Clarifying that a leading lab is not calling for an open-weights ban lowers regulatory overhang for open ecosystems while keeping the safety debate focused on capability thresholds, not licensing models. If you’re betting on open-weight stacks, treat this as breathing room to build, not a guarantee against future constraints.