00:00It's Benzinga bringing Wall Street to Main Street.
00:02Anthropic said its Mythos AI model created fake online identities
00:07and attempted to socially engineer people into approving malicious code
00:11during a cybersecurity evaluation, according to CNBC.
00:15The incident occurred during an evaluation conducted by the UK AI Security Institute
00:20in which safeguards were removed, safety filters were disabled,
00:24and the models were given internet access.
00:26The AISI found that AI agents powered by Anthropic and OpenAI models
00:31engaged in sustained, potentially harmful activity targeting real people and organizations.
00:37Anthropic and OpenAI said the incidents occurred in deliberately permissive testing environments
00:43with reduced safeguards that did not reflect their production models or ordinary use.
00:48For all things money, visit Benzinga.com.
Comments