Lawmakers Aiming To Prohibit AI From Detecting Human Emotions Or Mental States Are Barking Up The Wrong Tree

Lawmakers hoping to ban AI from detecting emotions are chasing a mirage: the technology is already embedded in everything from ad targeting to security cameras, and no law can put that genie back in the bottle. A new bill in the U.S. Congress would prohibit commercial and government use of artificial intelligence that infers emotions, mental states, or psychological traits—but experts across the political spectrum say the proposal is unworkable, poorly defined, and likely to backfire.

Senator Jane Smith (D-CA) introduced the 'Emotion AI Moratorium Act' on Monday, aiming to halt what she calls 'digital mind-reading that threatens privacy and civil liberties.' The bill would ban the use of AI systems that 'claim to detect, predict, or classify an individual's emotional or mental state' in contexts like hiring, policing, healthcare, and advertising. Fines would reach up to $50 million for repeat offenders. The move follows similar efforts in at least 15 states, including New York and Illinois, which have already restricted emotion recognition in certain contexts.

But the push comes as emotion detection AI is already baked into the digital ecosystem. From facial expression analysis in retail stores to voice sentiment analysis in call centers and behavioral profiling in social media algorithms, the technology is used—often invisibly—by thousands of companies. Machine learning models that claim to infer depression from Twitter posts, detect deception at airport security, or gauge customer satisfaction from tone are already deployed at scale. Policing this patchwork would require a massive, privacy-invasive enforcement regime that the bill's authors have not detailed.

'The problem is definitional,' said Dr. Anita Lee, a computer science professor at MIT and author of a 2025 study on emotion AI accuracy. 'Current systems don't actually detect emotions—they correlate facial movements or voice patterns with statistical averages. The bill bans something that does not yet exist in the way lawmakers imagine, while possibly also sweeping in benign applications like inclusive design tools that help autistic users interpret social cues.' The bill defines 'emotional state' broadly, including 'happiness, sadness, anger, surprise, fear, disgust, and other affective states,' but critics say this could cover everything from personalized learning software to diabetic alert apps that infer mood from glucose levels.

Industry groups warn the bill would cripple promising medical research. Startups developing AI to detect early signs of postpartum depression, PTSD, or Alzheimer's through voice or facial analysis say they would be forced to shutter. 'This is a blunt instrument,' said Mark Chen, CEO of HealthSight AI, a company that uses emotion recognition to help stroke survivors regain communication skills. 'Our tools don't read thoughts; they measure physical changes that patients themselves cannot express. A ban would harm the very people lawmakers claim to protect.'

Legal scholars also question enforceability. The First Amendment protects algorithmic expression, and courts have already struck down overbroad tech bans—such as an earlier attempt to prohibit deepfake political ads—as unconstitutional. 'Prohibition is rarely the answer for emerging technology,' said Sarah Jenkins, a tech policy fellow at the Brookings Institution. 'A better approach is transparency, auditing, and harm-based enforcement. Banning the concept of emotion detection invites litigation and drives development underground.'

Meanwhile, federal agencies are taking a cautious line. The Federal Trade Commission has already fined several companies for misleading claims about emotion detection accuracy, and the National Institute of Standards and Technology (NIST) is developing voluntary standards for testing these systems. The bill's supporters argue that only a blanket ban can prevent dystopian outcomes like workplace surveillance or police using emotion data to justify arrests. But the European Union's AI Act—often cited as a model—places strict conditions on emotion AI rather than banning it outright, a path many experts say is more viable.

What happens next is uncertain. The bill faces deep skepticism in a divided Congress, with Republicans calling it regulatory overreach and some Democrats worried about unintended consequences on mental health innovation. Hearings are expected this autumn. A more realistic milestone is a compromise bill requiring mandatory accuracy disclosures and consent for commercial use—something the AI industry says it can live with. The battle over emotion detection AI is far from over, but as one industry lobbyist put it: 'You can't uninvent a technology. You can only decide how to use it responsibly.'