Home / Technology / Anthropic AI's New Policy: Data Retention Changes
Anthropic AI's New Policy: Data Retention Changes
10 Jun
Summary
- Anthropic's new Fable 5 model has changed data retention policies.
- Mythos-class models retain prompts and outputs for 30 days.
- This update impacts existing Zero Data Retention agreements.
Anthropic has released Claude Fable 5, its most powerful public model to date. Alongside this launch, the company has implemented a notable change to its data retention policies for Mythos-class models.
Prompts submitted to and outputs generated by Fable 5 and other Mythos-class models will now be retained for 30 days. This retention is applied across all platforms offering these models and is intended for trust and safety purposes. This policy update was first highlighted by AI training company CEO Jun Park.
This new policy carries significant implications for Anthropic's enterprise and API clients. Lawyer Jessica Eaves Mathews points out that Fable 5 cannot operate under the Zero Data Retention (ZDR) agreements that other Claude models typically allow. Any existing ZDR agreements are superseded for traffic involving Fable 5 and future Mythos models. Organizations previously assured of data non-retention must now account for this mandatory exception.
