Can you imagine creating something so advanced that you are afraid to show it to the world? Well, that is exactly what has happened to Anthropic with its new model, Claude Mythos Preview. According to their announcement, this AI has such brutal programming capabilities that it could find and exploit software vulnerabilities better than almost any human cybersecurity expert, and it has already detected serious flaws in all the major operating systems and browsers on the market.
In light of this, they have decided not to make it public for now. Instead, they have launched the ‘Project Glasswing’, an initiative in which they collaborate with giants like Apple, Google and Microsoft to use this model exclusively for defensive purposes. Anthropic is earmarking up to $100 million in usage credits so that more than 40 organizations can scan and secure critical infrastructure before these capabilities fall into the wrong hands.
It is a fascinating and somewhat unsettling move, to be honest. It shows us how fast artificial intelligence is advancing and the security challenges we face. It seems we are entering a new era where AI not only helps us draft emails, but could be the ultimate weapon both to attack and to defend our digital systems.
