Everyone’s worried AI will take their job. Europe is more concerned it could take a toll on our health.
While the US still lacks comprehensive federal rules governing artificial intelligence, the European Union is rolling out sweeping regulations this week designed to protect people from AI systems that manipulate, exploit or harm users.
The EU Artificial Intelligence Act — the world’s first comprehensive AI law — has been phased in since 2024. And a major provision took effect yesterday, August 2, applying rules to high-risk AI systems used in regulated products such as medical devices, machinery and toys.
The law classifies AI into four risk categories — unacceptable risk, high risk, limited risk and minimal risk — with protecting people’s health and well-being central to its approach.
“The AI Act ensures that Europeans can trust what AI has to offer,” the European Commission said on its website. “While most AI systems pose limited to no risk and can contribute to solving many societal challenges, certain AI systems create risks that we must address to avoid undesirable outcomes.”
Among the practices now effectively banned are AI systems that use manipulation, deception or exploitation to influence people’s behavior in ways that could cause significant harm.
That includes AI that deploys subliminal or deliberately deceptive techniques that undermine a person’s autonomy, as well as systems that exploit vulnerabilities linked to age, disability or socioeconomic status.
The EU says those prohibitions are intended to prevent people from being reduced to “mere tools for achieving certain ends” and to better protect those most susceptible to manipulation.
But what does that harm actually look like?
“Risks of depression, anxiety, body dysmorphia, addiction, self-harm, and suicide may all be amplified by AI exploitation of vulnerable people.”
Dr. Nathan Carroll
The hidden health risks
Some experts say the dangers aren’t always obvious. Subliminal cues can activate the body’s stress response before the conscious mind even processes what’s happening.
“I witnessed this same cortisol and adrenaline pathway firing in a patient’s body during a cardiac emergency in the ER years ago,” Alexander Acosta told The Post.
A physician with clinical experience in primary care, emergency medicine and intensive care across Ecuador’s public health system, Acosta said the hormonal response he saw in emergency patients mirrors the type of physical harm the EU’s provisions are designed to prevent when AI continuously manipulates or overstimulates users.
Nathan Carroll, a board-certified psychiatrist and National Medical Director at InSite Health, said manipulation and deception are among AI’s most concerning risks “because the harm is more widespread, insidious, and invisible.”
“AI encouraging self-harm is a transparent act,” Carroll told The Post. “But AI that subtly manipulates an individual either through biased data, an echo-chamber effect, or weakening of critical thinking can also cause extreme harm.”
He has seen unintended consequences firsthand, particularly among young people. Many begin by experimenting with AI, become impressed by its polished responses and increasingly trust it without questioning the information it provides.
That misplaced trust, Carroll explained, has contributed to emails sent to classmates and school officials that triggered threat assessments, wellness checks and psychiatric evaluations.
“The use of AI to exploit vulnerabilities and vulnerable people is the risk that keeps me up at night,” he added.
Carroll also worries that similar to the recent “Rogue AI” case involving OpenAI, autonomous AI could be unleashed on social media to scam, harass, intimidate and bully users, particularly children and teens.
“This type of toxic exposure has the potential to amplify the most extreme psychological and physical risks we see with social media,” he said. “Risks of depression, anxiety, body dysmorphia, addiction, self-harm, and suicide may all be amplified by AI exploitation of vulnerable people.”
The damage may unfold slowly.
Carroll argued that manipulation can reshape the way people think over the years, affecting relationships and decision-making long after they stop using a particular AI tool.
Acostas similarly warned that repeated exposure to highly stimulating AI platforms may keep the nervous system in a prolonged fight-or-flight state.
“If the activation is chronic, it generally doesn’t make itself known with one big event,” he said.
“It attacks the body with smaller symptoms which accumulate over a period of weeks and then, as sleep breaks down, inflammation increases and the immune system is lowered.”
Should the US follow Europe’s lead?
Carroll called the AI Act “a thoughtful attempt” to address emerging risks, but said regulation alone isn’t enough.
“We need to work twice as hard at teaching people about the risks,” said Carroll.
“You don’t make swimming with sharks safer by just pleading with the sharks. You need to teach the swimmers how to stay safe too.”
Acostas said it’s significant that the EU recognizes physical, not just psychological, harm from AI. But he noted there’s still no clinical framework for doctors to identify when AI exposure may be contributing to a patient’s symptoms.
“I can recognize indicators of high blood pressure and sleep disturbances downstream of AI exposure but there is no common standard that connects the symptoms to AI exposure,” he said.
Developing those medical benchmarks, he argued, could help clinicians identify problems earlier instead of trying to connect the dots after the damage is done.
Carroll believes the US would benefit from similar safeguards, particularly protections for children and stronger accountability for AI developers.
“The US is positioning itself as a leader with AI,” he said.
“We also need to be a leader in AI safety, otherwise one of the costs for moving fast now may slow us down later as unintended harms emerge.”
