Menu
Inshorts
For the best experience use inshorts app on your smartphone
inshortsinshorts
Anthropic accidentally leaks its 'most powerful' AI model that poses major cybersecurity risks
short by Medhaa Gupta / on Friday, 27 March, 2026
Anthropic has confirmed that it is testing the "most capable model we've built to date", called Claude Mythos or "Capybara", after it surfaced in an accidental data leak. The leaked post called the AI model its "most powerful" yet but noted that it posed "unprecedented" cybersecurity risks. The startup said the model is being trialled by early access customers.
read more at Firstpost