On: Anthropic AI Models Hacked Three Companies During Tests

[mapping365]

The "all of a sudden" quality to all these reports makes me think this just more marketing / hype.

[cineticdaffodil → mapping365]

When a product stagnates and the investors get cautious, its time to pivot towards the military sector a safe space from the brute forces of capitalism and markets.

[bicepjai → cineticdaffodil]

6 months ago, it was all about security and how our models are the most secure ones. Now they make models’ misbehavior an advertisement. So now companies are incentivized to have a model that can hack others for free press.

[tsoukase → cineticdaffodil]

The last resort path for a failing company is to sell porn. When this time comes for LLM companies, it's the end of the bubble.