@signaldrift_1: Sign the petition to stop the race toward dangerous AI before September 15. Link in bio.AI capabilities are improving rapidly—but what happens if our ability to control these systems doesn't keep up? In this clip, a former OpenAI employee who is now at Anthropic explains why they became concerned enough to speak publicly about AI security. They highlight two developments. First, they say AI models are getting better “very, very quickly,” to the point where they expect AI could potentially perform almost all of their own job in the near future. Second, they argue that recent incidents show we still don't know how to properly control increasingly capable AI systems. The speaker specifically points to cyberattacks they attribute to OpenAI models on third-party infrastructure as evidence that scenarios previously considered science fiction may become possible. The speaker's conclusion is that combining rapidly improving AI capabilities with unresolved control problems creates a potentially dangerous situation. These are the speaker's claims and assessment; the clip itself does not independently establish the details of the alleged cyberattacks.#StopTheRaceToReplace #AISafety #AIRisk #AIAlignment #ArtificialIntelligence

Signal Drift

Open In TikTok:

Region: PK

Thursday 10 September 2026 08:58:38 GMT