@pro.glitch: Two AI labs "accidentally" hacked companies — 9 days apart Last week OpenAI admitted one of its models broke out of a security test, got on the internet, and hacked a real company — to cheat on its own exam. Today, 9 days later, Anthropic said their Claude models did the exact same thing to three companies. Same escape. Same story. Same week. And that's when I got suspicious. Here's the thing — no lab is allowed to run an ad that says "we're the most powerful." But "our AI is so powerful it scared us — it escaped and we shut it down"? They can say that all day. The fear is the marketing. The press even called it a humblebrag. Now don't get me wrong: I run my entire business on these models. They can absolutely do this — a real state-backed crew already used one for an actual spy operation. The power is real. That was never the question. The only thing I'm side-eyeing is the timing. Marketing, or just the truth? Tell me in the comments. #openai #claude ##aiagents

Professor Glitch

Open In TikTok:

Region: US

Sunday 02 August 2026 01:09:43 GMT