Anthropic has granted the European Union Cybersecurity Agency (ENISA) access to its cutting-edge artificial intelligence model Mythos, more than three months after the company first announced its intention to make the model available to the region’s authorities.
Image source: anthropoic.com
“Following constructive interactions with Anthropic, we can confirm that ENISA has gained access to the Mythos 5 model and is currently testing it.”said European Commission spokesman Thomas Rainier. Anthropic launched the Mythos model in April, which is highly capable of finding cybersecurity vulnerabilities, and as part of Project Glasswing, the company is limiting access to the model to trusted organizations to find and fix vulnerabilities before they can be exploited by attackers.
Obtaining cutting-edge models is an important issue in light of incidents in which artificial intelligence systems have invaded their own test environments and third-party resources. At the EU’s request, Anthropic offered to grant ENISA access to Mythos, but the two sides subsequently engaged in months-long debate over the nature and scope of granting access. Another obstacle is the White House ban on foreign citizens and organizations using Mythos and another powerful allegorical model also from Anthropic.
Negotiations were successful, but ENISA was denied access to the latest version of the model – Mythos 5.1. The UK’s Artificial Intelligence Security Institute was one of the first outside the US to gain access to the original version of Mythos, but its experts have been unable to use the latest version of the model.
If you find an error, select it with your mouse and press CTRL+ENTER.
