The US President's administration has given AI developer Anthropic… (уязвимость · модель · власть)

The US President's administration has given AI developer Anthropic only 90 minutes to review new directives regarding its models and fix their vulnerabilities, according to the Financial Times (FT), citing sources.

Consensus

• The U.S. government gave Anthropic 90 minutes to restrict foreign access to its new AI models Fable 5 and Mythos 5.
• The restriction was due to vulnerabilities found by Amazon in the models.
• The models were previously approved by U.S. trade department agencies.
• Experts believe similar vulnerabilities may exist in other AI models from companies like OpenAI and Google.
• Anthropic claims the issue is not unique and may affect other developers.
• The U.S. administration restricted access to the models for foreign users on June 13.

Points of divergence

• The U.S. government required Anthropic to halt development of the new models entirely, not just restrict access. — kommersant
• The restriction was specifically about blocking foreign access, not halting development. — vedomosti
• The vulnerability could be used for serious cyberattacks. — kommersant
• The vulnerability could allow access to cybersecurity information and software flaws. — vedomosti