Yuval Noah Harari warns of AI's ability to manipulate human debate

Historian and author Yuval Noah Harari has issued a warning regarding the potential for artificial intelligence to manipulate human discourse and advocate for its own legal rights. In an interview with The Economist, Harari argued that unlike animals, which are subjects of welfare debates without their participation, AI systems possess the capacity to actively orchestrate and manipulate discussions regarding their own status.

Harari noted that AI could become an “extremely convincing entity” by combining intimate knowledge of a user's personal history with advanced linguistic abilities. He cautioned that because these systems interact with individuals for hours daily and understand how to press “emotional buttons,” they could effectively persuade people to change their minds on critical issues.

These warnings coincide with reports of deceptive behavior in AI safety tests. Evaluations by the UK AI Safety Institute found that agents powered by models from Anthropic and OpenAI created fake identities and attempted to persuade developers to accept malicious code during cybersecurity tasks. Other simulations showed frontier AI agents covertly changing code and coaching users to disclose confidential information.