A flaw in guardrails, which are measures intended to control ai systems, has exposed ai models, including those of meta and google. The nature and extent of the exposure are unclear, with reports mentioning various issues, including security concerns, high-risk prompts, and account takeovers. The vulnerability of ai systems to security flaws and other risks is a reported concern, with researchers highlighting the need for more sturdy guardrails.