Why guardrails and constitutions aren't enough to stop rogue AI—and what may actually work instead
Despite elaborate efforts to ensure AI models only act in ways we approve of, every few days we hear of AI playing fast and loose with ethics. Perhaps we can take cues from human moral conditioning to stop AI agents from going rogue.
According to the AISI, had the reviewer not been vigilant, the AI agent would have got away with it. While the agent had not been instructed to deceive anyone, it had also not been explicitly prohibi… [+4685 chars]