Updated
Updated · Bloomberg · Aug 20
Anthropic Lets Business Customers Keep 30-Day AI Data on Their Own Cloud
Updated
Updated · Bloomberg · Aug 20

Anthropic Lets Business Customers Keep 30-Day AI Data on Their Own Cloud

3 articles · Updated · Bloomberg · Aug 20

Summary

  • Later this year, Anthropic plans to let enterprise users of its most advanced Claude models satisfy a 30-day data-retention rule on their own cloud infrastructure instead of Anthropic’s systems.
  • The shift softens a June policy covering Mythos, Fable and future frontier models, which required Anthropic to hold all customer data for 30 days to detect and block novel cyberattacks.
  • More than 100 customers in highly regulated industries have been helping shape the new option, which Anthropic had been developing for months before public criticism emerged.
  • OpenAI on Wednesday began testing a rival zero-retention approach called private safety processing with Databricks and Microsoft, with a broader release planned for September.
  • Anthropic had warned the original retention policy could be unpopular with customers accustomed to zero retention and could hurt its business if competitors did not match it.

Insights

As AI giants shift safety storage to clients, are they solving privacy concerns or just offloading infrastructure costs?
Will letting enterprises control their AI data storage create new blind spots for detecting advanced cyber threats?