Anthropic to share Mythos cyber flaw findings with global finance watchdog

Anthropic will brief the Financial Stability Board on the implications of its Claude Mythos AI model for cyber defenses. The model has raised concerns among experts due to potential misuse by hackers. Anthropic has not released Mythos publicly, citing advanced capabilities that can highlight previously unknown flaws in IT systems. Instead, it has provided access to selected technology companies and banks, including Apple and JP Morgan, to help identify weaknesses the model might locate. The UK AI Security Institute reviewed a version shared with banks and tech firms and reported a notable capability jump. It said the latest version solved the previously unsolved “cooling tower” cybersecurity test in three of ten attempts, a first for any model it tested. The institute also noted rapid progress in autonomous cyber task completion and plans tougher future tests.

"Anthropic has declined to release Mythos publicly, after announcing that it had advanced capabilities in highlighting previously unknown flaws in IT systems which could be utilised by hackers. Instead, the company has given access to the model to a group of tech companies and banks, including Apple and JP Morgan, to help them identify any weaknesses that Mythos might locate."

"The UK's AI Security Institute (AISI), which assesses advanced AI models, issued an updated appraisal of Mythos last week after scrutinising the version that has been released to banks and tech companies. It said the latest iteration it had seen represented a notable capability jump even on the preview version it had tested the previous month."

"AISI said the latest version of Mythos had completed a previously unsolved cybersecurity test, dubbed cooling tower, in three out of 10 attempts. This was a first for any model tested by AISI. Frontier AI's autonomous cyber and software capability is advancing quickly: the length of cyber tasks that frontier models can complete autonomously has doubled on the order of months, not years, AISI said."

"The FSB plan, which was first reported by the Financial Times, was confirmed by a source familiar with the discussions between the regulator and Anthropic. The FSB monitors and makes recommendations about the global financial system and includes officials from leading economies including the US, the UK, Australia and China. Its steering committee includes senior central bank and finance ministry officials."

Read at www.theguardian.com