Japan to receive early access to Claude Mythos AI model from Anthropic
Officials say US plans could let Japanese authorities and firms prepare defences against an advanced AI able to spot system weaknesses.
- The Japanese government and companies will receive early access to Anthropic's Claude Mythos AI model within two weeks.
- The model is not yet publicly available and is known for its powerful ability to identify system vulnerabilities.
- This access is intended to help Japan preemptively strengthen its cybersecurity, though concerns exist that the model could also be exploited for cyberattacks.
- Japan's Financial Services Agency has urged financial institutions to treat the model's arrival as a company-wide management issue.
Japanese Finance Minister Satsuki Katayama said Friday (May 22) that US Treasury Secretary Scott Bessent had informed her of plans to give the Japanese government and Japanese companies access to the Claude Mythos artificial intelligence model within two weeks.
Katayama said the plan was conveyed during her May 12 meeting with Bessent. The model was developed by US startup Anthropic and remains unavailable to the general public, with access currently limited to some US companies and groups.
Mythos is known for its strong ability to identify system vulnerabilities, a feature that has raised concerns that it could be exploited for cyberattacks. At the same time, early access is expected to help Japan take pre-emptive steps to strengthen cybersecurity.
Also on Friday, the Financial Services Agency urged Japanese financial institutions to move quickly in responding to advanced AI models, including Mythos. The agency stressed that top management officials should treat the matter as a company-wide management issue.
The FSA also said proactive system shutdowns should remain an option if even thorough protective measures are unable to prevent cyberattacks.
[Copyright The Jiji Press, Ltd.]
