Myth is a concern for the company. The rest are lulled by false cybersecurity

A few weeks ago I wrote in these pages that we have crossed the Rubicon in cyber risk: that the threshold between human attacks and machine attacks has been crossed by Anthropic’s mythos. That the model would not remain under control. That OpenAI was close behind.

Within two weeks, on the same day the model was announced, Anthropic confirmed that it was investigating reports that unauthorized users had accessed Mythos Preview through a third-party portal. The containment on which the entire Glasswing access model was based had failed at the starting gate. At the end of last week, the British government’s AI Security Institute confirmed the rest: border laboratories are running at full speed and capacity is increasing.

Load…