Either Claude Opus 5 just leaked a trove of private 2024–2025 chat logs from Anthropic employees (& general public) asking the AI to edit their personal letters, or it’s generating one hell of a hallucination. It’s almost certainly the latter. But read the examples and you might start second-guessing.
It leaves you wondering what exactly is going on with Claude Opus 5—and whether we’re looking at a hallucination, or a model trained on its own creator’s internal data.
4 Comments
solrize@lemmy.ml · 5 pts · 16d
Who and who?
decolo@piefed.social · 5 pts · 16d
Anthropic leadership
otter@lemmy.ca · 1 pts · 16d
Added more context to the title
Zwuzelmaus@feddit.org · 3 pts · 16d
No problem, as long as it doesn't tell my beautiful brain's internal data 😇