Anthropic just handed the EU its unreleased 'cyberweapon' after hours of emergency talks

Project Glasswing, Anthropic's unreleased 'cyberweapon', is home to a previous model of Claude Mythos and offers unprecedented hacking tools to anyone using it, and the AI giant has just handed over the keys to the European Union following hours of emergency talks.

Many of the world's leading AI figures have been clear about the tech's potential for good across countless disciplines, but one of the biggest concerns relates to its role in the world of cybersecurity, as it could provide game-changing tools on both sides of the table.

OpenAI CEO Sam Altman has previously predicted that there will be a 'world-shaking' cyberattack this year with AI at the heart, and something that could perhaps be key to achieving that and preventing it at the same time is Anthropic's Project Glasswing.

Anthropic can't achieve it all on its own, however, and that's why it's enlisted the help of the European Union Agency for Cybersecurity (ENISA), as together they have already identified countless zero-day exploits and even a vulnerability in OpenBSD that has been missed for over 27 years.

Anthropic has partnered up with the European Union's cybersecurity team to provide access to Project Glasswing (Nicolas Tucat/AFP via Getty Images)

As reported by TechZine, the access granted to ENISA is still restricted, limiting the amount that the EU can play around with Claude Mythos' industry-shattering tools, but this partership could prove vital to the success of Project Glasswing when it is widely released in the near future.

Further reports have indicated that over 10,000 critical software flaws have been identified by Claude Mythos in the first month after its release, and Anthropic has reported an exploit success rate of around 72.4 per cent, which is night and day compared to the virtually nonexistent rate of the previous model.

While Project Glasswing has been created alongside the help of many of the world's largest tech companies including Google and Microsoft, alongside Apple and Amazon, Anthropic has also given access to the US government so that it can identify and remove any potential vulnerabilities that crop up within official systems.

Anthropic is simultaneously working with the US government on Project Glasswing, identifying further dangerous exploits (Samyukta Lakshmi/Bloomberg via Getty Images)

Any official partners are only allowed to access and use Claude Mythos for cybersecurity and protective measures, but the dangers are understandably present if it ever got into the hands of anyone who had a more nefarious motive.

There's arguably no better tool for defeating cybersecurity protections than the software used to set them up in the first place, and evolutions spawning from that would undoubtedly cause chaos across the digital world.

That's why collaborations between Anthropic and the world's leading political organizations remains a vital step in security our cybersecurity future, as it not only helps uproot any exploits that currently exist, but hopefully prevents the tool from falling into the wrong hands going forward.