I watched a video by SomeOrdinaryGamers called The Most Dangerous AI Just Went Live, and it is something bigger than just another model release. This is not a chatbot update. AI is moving out of its novelty phase and into something serious.
Anthropic announced Project Glasswing on April 7, built around a gated model called Claude Mythos Preview. Anthropic says Mythos found thousands of significant software vulnerabilities, and instead of releasing it broadly, it is limiting access to select partners for defensive cybersecurity work. Those partners include names like Amazon, Microsoft, Apple, Google, Nvidia, CrowdStrike, and Palo Alto Networks, a very exclusive list.
The real money and real importance of AI ends up being in security, infrastructure, and systems businesses cannot afford to have failed. Anthropic has been discussing Mythos’s offensive and defensive cyber capabilities with U.S. government officials. This is not a toy being marketed to everyone. It is a capability being controlled because open access to it would be risky. Watch SomeOrdinaryGamers video and see what people can already do with existing open source models, if Mythos or a similar model becomes open source there is no telling what people will do.
Mythos reportedly identified problems across every major operating system and web browser, and even autonomously developed exploits in testing. The question now becomes, how much do we trust these models in the hands of the companies that receive them?
AI is becoming more serious really fast. We are now past the flashy stuff and asking about security and safety, even human lives.
What do you think about AI now being able to basically take over the internet?
Discover more from Shayan Hosseini
Subscribe to get the latest posts sent to your email.
