日本大型银行恢复使用 Anthropic Claude Mythos AI 模型权限

TOKYO -- Top Japanese banks have regained access to Anthropic's powerful Claude Mythos artificial intelligence model, Nikkei learned Thursday, after the U.S. government had ordered overseas access suspended in June over cybersecurity concerns.
The U.S. AI company unveiled Mythos in April. Citing the high potential for misuse, including in cyberattacks, Anthropic initially limited access to a select group of users.
Japan's top banks were expected to receive access, but uncertainty had arisen after U.S. President Donald Trump's administration told the company to suspend the model's availability to all foreign nationals.
Among other advanced AI models, OpenAI has granted access to its top models to big Japanese banks, which also have access to a Google-developed product that can identify system vulnerabilities and automatically generate software patches to address them.
Combining multiple advanced AI systems could improve the ability to detect vulnerabilities. The banks are expected to begin evaluating whether Mythos can be deployed for vulnerability detection and other cybersecurity applications.
Masahiko Kato, chairperson of the Japanese Bankers Association and president of Mizuho Bank, said at a June news conference that, in order to prepare for AI-driven cyberattacks, "there could be situations where we proactively suspend certain services, such as ATMs and internet banking."