Anthropic customers restrict Fable use over data retention concerns

Anthropic’s launch of Claude Fable 5 kicked off a corporate trust crisis that’s still reverberating across enterprise AI adoption. The culprit isn’t the model’s capabilities. It’s a mandatory 30-day data retention policy that came bundled with Fable 5’s release on June 9, 2026, effectively overriding the zero data retention agreements that enterprise customers had previously relied on. What changed and why it matters Claude Fable 5, categorized internally as a “Mythos-class” model, represents a new tier of AI capability at Anthropic. With that capability came a policy shift: all prompts and outputs submitted to Fable 5 would be stored for 30 days, regardless of existing enterprise agreements. Anthropic framed the change as a safety necessity. The company argued that monitoring for potential misuse, cyber threats, and bypass protocols required retaining data for a minimum period, something that simply wasn’t possible under zero data retention frameworks. That rationale landed poorly with customers in finance, law, and healthcare, sectors where data confidentiality isn’t a preference but a regulatory requirement. Microsoft moved fastest. On June 10, just one day after Fable 5 went li...