Anthropic Revamps Data Policy Introducing ‘Enterprise Frontier Safeguards’
AI safety firm Anthropic unveiled on Tuesday, September 1, a significant revision to its data retention policy for business clients. This change, driven by extensive feedback from enterprise users, marks a pivotal shift in the way the Claude chatmaker manages sensitive corporate data.
The company is introducing a new system, Enterprise Frontier Safeguards, to replace its existing policy. This system will provide businesses with enhanced control over the review, storage, and management of their data. Importantly, the new framework will enable companies to conduct automated safety monitoring without necessitating human review by Anthropic employees. This feature addresses a major concern voiced by enterprise clients in highly regulated sectors. Anthropic has also confirmed that there will be no additional charges for these new safeguards. The controls will be applicable to both direct Anthropic customers and those accessing Claude via cloud providers.
The controversy originated from a policy implemented in June, coinciding with the launch of Anthropic’s most sophisticated AI models, Claude Fable 5 and Mythos 5. This policy mandated a 30-day retention of all traffic from these models to bolster defense against cyberattacks. Kate Jensen, Head of Americas at Anthropic, informed CNBC that the company devoted hundreds of hours collaborating with over 100 customers, including Salesforce, to devise an alternative. The Enterprise Frontier Safeguards will be rolled out in stages, with wider availability aimed for fall 2026. In addition, Anthropic announced the release of two updated models, Claude Fable 5.1 and Mythos 5.1, boasting enhanced capabilities for programming, research, and knowledge work.
Source: CNBC – Anthropic Adjusts Data Retention Policy in Response to Customer Feedback