Published on
Anthropic is expanding testing of Mythos, its latest cybersecurity model, by bringing in around 150 additional partners.
The company, seen as the main rival to OpenAI’s ChatGPT, announced in April that it was launching Mythos, a model it said “has already found thousands of high-severity vulnerabilities, including some in every major operating system and web browser.”
Mythos was initially restricted to 50 initial partners who had access to the preview model under the so-called “Project Glasswing” and, according to Anthropic, have used it to find more than 10,000 high or critical security flaws.
Now, the project includes organisations from more than 15 countries and covers many industries, including power, water, healthcare, communications and hardware, the company said.
“What each partner has in common is that a successful attack on their codebase could be catastrophic,” the company said, estimating that a cyberattack would affect more than 100 million people for most of their new partner organisations.
One potential customer is the European Commission, which confirmed to US media that Anthropic had approached it to use Mythos. Euronews Next is working to independently confirm this.
The company is expanding their programme because it estimates that in the next year, more AI companies will have similarly powerful models and “could release them without safeguards that prevent misuse.”
“In that world, cyberattacks could occur much more often, and in much more unpredictable forms,” the company wrote.
The expansion of Project Glasswing comes just a day after Anthropic filed confidentially for an initial public offering. The company also recently received $65 billion (€59bn) in funding and is now worth nearly $1 trillion (€878bn).