“Reliably assess potential threats.” Anthropic and Brussels are getting closer to an agreement
According to the official, the company issued the official invitation after a meeting with EC representatives in San Francisco last Thursday. He added that the European Union must now work out a mechanism enabling safe access to the modelwhile maintaining appropriate security.
Bloomberg reported on Monday that ENISA, the EU's cybersecurity agency based in Athens, is expected to gain access to Mythos.
European Commission spokesman Thomas Regnier said in a statement that the Commission held a meeting “a series of fruitful meetings with Anthropik” and “welcomes the latest developments regarding future access opportunities.”
“Let's not forget that Mythos is not a single case”
Anthropic unveiled Mythos in early April, warning that this model outperforms most humans in detecting and exploiting cybersecurity vulnerabilities. This raised concerns that the tool could be used for mass attacks on key and sensitive systems if it ends up in the wrong hands of cybercriminals.
“This latest step is critical to reliably assess potential threats,” Regnier said. — Let's not forget that Mythos is not a single case. A whole new wave of powerful models is now appearing on the market.
An ENISA representative informed that the agency does not yet have active access to the tool, but is working on its implementation.
The Commission is preparing a formal action plan to respond to the emergence of advanced AI hacking tools. According to an industry representative the institution plans to publish it before the summer break.
