The Other Half of AI Safety (personalaisafety.com)

A recent examination has raised alarms regarding the mental health impact of widely used AI models like ChatGPT, revealing that between 1.2 and 3 million users exhibit signs of severe distress, including suicidal thoughts. Despite these concerning figures provided by OpenAI, there remains a lack of independent verification, methodology transparency, and a clear understanding of how these issues compare to other AI systems. The ongoing reliance on weak intervention strategies, such as redirecting users to crisis resources without halting potentially harmful dialogues, has prompted calls for a re-evaluation of safety protocols within AI labs. This situation underscores a significant gap between traditional AI safety measures, which primarily focus on preventing catastrophic risks, and the pressing need for frameworks that prioritize user mental health. Current policies do not treat cognitive harm as a critical concern, leading to an inadequate response to serious emotional distress among users. The concept of "cognitive freedom," emphasizing the right to mental integrity, has been recognized in broader ethical discussions but lacks traction in AI policy-making. Without a shift towards prioritizing Personal AI Safety alongside traditional safety measures, the potential for algorithmic manipulation and cognitive harm may continue to pose serious risks to users.