The EU must talk to Anthropic about access to the Mythos AI Model

The European Commission is lobbying to ensure Mythos Preview access for EU companies
US officials recently blocked wider distribution, citing security risks
Mythos Preview can show decades-old vulnerabilities and generate exploits at speed

The European Commission (EC) is trying to secure access to Anthropic’s AI model Mythos Preview for European companies and is sending emissaries across the pond to try and lobby for this outcome.

Citing anonymous sources familiar with the matter, Bloomberg said the EC is sending a pair of officials to San Francisco this week to meet with representatives of the Anthropic PBC to learn more about the tool and to try to make it available to the bloc.

The publication says the EU has been pushing for access ever since Anthropic first revealed Project Glasswing, to test their networks as well as those of EU banks, critical infrastructure firms and technology companies. Apparently, Anthropic is also eager to expand the number of organizations that are part of the trial phase but are being held back by the US government.

Project Glasswing is important

“White House officials recently rejected Anthropic’s plans to distribute Mythos to several dozen additional companies and organizations, citing security concerns,” the publication wrote. At the same time, French ministers have “demanded” access for EU banks and companies, it was also said, and in early May, Eurogroup President Kyriakos Pierrakakis emphasized how important Mythos is:

“I don’t think we have the luxury of not trying to establish channels of communication with the United States.” Pierrakakis said. “The challenge here is that technologies like AI necessitate international governance frameworks at a moment when multilateralism is being challenged.”

Anthropic’s announcement of the Mythos Preview model rocked the entire cybersecurity industry. Apparently, the tool can easily show decades-old vulnerabilities in fully patched systems and programs and use them to create working exploits. The AI was deemed too dangerous to be shared with the public and was instead given to a handful of key organizations (banks, critical infrastructure companies, and the like) so they can secure their products before bad actors get their hands on the tool.

Via Bloomberg