00:00It's Benzinga, bringing Wall Street to Main Street.
00:02The U.S. government is unlikely to extend AI export controls tied to Anthropix models to
00:07other companies, the information reported Saturday. The administration's concerns center
00:12on security vulnerabilities in Anthropix Fable 5 and Mythos 5 models. The government said Fable
00:19could still find vulnerabilities in sensitive systems despite public release guardrails
00:23and cited risks of advanced AI capabilities spreading to foreign adversaries.
00:28Both Mythos class models were disabled following a government order restricting access to foreign
00:33nationals. Anthropix argued in a blog post that similar vulnerabilities exist in other publicly
00:39available models, including recent OpenAI releases. David Sachs, co-chair of the President's Council
00:45of Advisors on Science and Technology, said Anthropix CEO Dario Amadei was asked to fix the
00:51vulnerabilities or de-deploy the model, but declined. For all things money, visit Benzinga.com.
Comments