Post #3316765

so what I need everyone to know is that Anthropic unilaterally changed its terms effective yesterday for anyone using one of its new Mythos-class models (which they also added to the platform unilaterally, and they’re very hard to turn off). even if you contracted with them for zero retention, zero model training, they’re overriding that and will retain any queries and their output using Mythos models to be reviewed and assessed by their tools and staff for safety for 30 days. they don’t say this explicitly but I assume they will also use them to train their model, at least in artifact form. I’m not telling you what to do but I wouldn’t put anything confidential or privileged into a prompt using these models for response. https://support.claude.com/en/articles/15425996-data-retention-practices-for-mythos-class-models