Anthropic Launches Claude Fable 5.2 and Mythos 5.2, Amid Report of Korean Corporate Account Leak

At a glance

Anthropic has released two new AI models with better coding, reasoning, and robot control. At the same time, reports came in that information from about 100 Claude accounts used by Korean companies had leaked.

Key points

1. 1

   Fable 5.2 is open to general users, while Mythos 5.2 can only be used by verified organizations in cybersecurity and life sciences.

2. 2

   On the FrontierCode 2.0 coding benchmark, Fable 5.2 scored 49.1%, higher than the previous model.

3. 3

   The leak was traced to a vulnerability in external software that Anthropic uses. The company said chat content and payment information were not accessed.

4. 4

   The Personal Information Protection Commission is reviewing the reports in advance and has not yet decided whether to start a formal investigation.

Why it matters

As AI models get faster, how companies protect the data they entrust to these systems is becoming just as important.