AI Has A Darker Side In Healthcare, Expert Says

TL;DR

For listenersOffer from Amazon

Turn quiet afternoons into listening time

• Thousands of audiobooks, podcasts and originals
• Listen on your phone, tablet or Echo — also offline
• Cancel anytime

Try Audible free →Free trial for new members

As an affiliate, we earn on qualifying purchases.Hide

At the Association of Health Care Journalists Fall Summit in Los Angeles, UC Berkeley researcher Mishal Khan warned that AI in healthcare may harm workers even when it does not eliminate jobs. She cited risks including de-skilling, monitoring, loss of autonomy and pressure on wages, while another researcher described medical residents struggling to interpret EKGs without AI support.

AI tools may create workplace harms in healthcare even when they do not replace workers, Mishal Khan, a senior researcher at the UC Berkeley Labor Center, said at the Association of Health Care Journalists Fall Summit in Los Angeles. Khan pointed to concerns including workers losing clinical skills, having their work monitored and facing reduced control over their jobs.

Khan grouped workplace uses of AI into electronic monitoring, algorithmic management and task automation. Monitoring can involve cameras tracking workers, while algorithmic management can analyze conversations and prompt employees on how to respond. Task automation shifts specific duties to software. These examples were presented as broader workplace patterns, not as evidence that every healthcare employer uses such systems.

In her research on AI’s potential effects in healthcare, Khan identified de-skilling as a particular concern: workers may lose specialized abilities when automated tools take over tasks. She said this could affect healthcare workers’ scope of practice. Other risks she listed included job loss, theft of work product, lower wages, work intensification, injuries, discrimination and reduced ability to organize.

Tina Hernandez-Boussard, a Stanford University professor of biomedical informatics who spoke in another summit session, described medical residents sent to help at a hospital cardiac ward in Uganda. After the facility lost electricity, there was no AI system available to read EKGs. Hernandez-Boussard said the residents were not confident synthesizing the information to make recommendations. The account illustrates a concern about dependence on automated support, but the report did not provide further details about the residents, the hospital or the clinical outcome.

Risks to Skills and Worker Control

The concerns go beyond whether AI directly replaces a job. If workers rely on automated systems for tasks they once performed themselves, clinical skills and confidence may weaken, potentially affecting their ability to work when systems are unavailable. The Uganda example raised that possibility, though it does not establish how widespread such effects are.

Monitoring and algorithmic management also raise questions about privacy, autonomy and accountability. Khan said workers have concerns about what information ambient scribe systems collect about them and whether that information could later be used in employment decisions. She also raised the question of who is responsible when an AI recommendation contributes to an error. The report did not identify a specific case or legal ruling resolving that issue.

For patients, working conditions matter because they can shape how care is delivered and how much time and attention workers can give to patients. Khan said automated handoffs could contribute to alienation or detachment from patient relationships. These are concerns raised by the researchers, not measured outcomes established by the report.

Workplace AI and State Responses

Khan’s remarks placed healthcare AI within a wider debate about how workplace technology can change the balance of power between employers and workers. She said digital technology has been used for decades to increase labor extraction and worsen working conditions. Her framework distinguishes different mechanisms: tracking workers, directing their interactions, or automating parts of their jobs.

The report also described several state-level actions cited by Khan. She said laws in Illinois and Vermont prohibit replacing mental health workers with AI, and that a California law bars companies from advertising an AI chatbot as a health professional. Khan also said a California bill under consideration would prevent workers from being required to train their replacements, including AI systems. The report did not give bill names, dates or further legislative details.

Khan cautioned against treating technology as inherently beneficial or harmful. She said AI may benefit workers, but argued that without meaningful worker power, harms can outweigh benefits. Her comments focused on labor effects and safeguards, rather than evaluating the clinical accuracy of AI systems generally.

“Even if they don’t lose a single job to AI, it doesn’t mean there are no harms to workers.”

— Mishal Khan, senior researcher at the UC Berkeley Labor Center

Evidence and Liability Questions Remain

The report does not quantify how many healthcare workers are affected by AI monitoring, algorithmic management or automation, nor does it establish how often these systems cause de-skilling, wage losses or patient-care problems. Khan’s research findings and the Uganda account raise concerns, but the report provides no study methods, outcome data or comparison group with which to measure the scale of those effects.

It is also unclear which ambient scribe tools collect worker data, how employers use that information, or what protections apply in specific workplaces. Khan posed liability for errors involving AI recommendations as an open question; the report did not describe a particular incident or provide a legal determination. Legislative details and the current status of the California bill were not specified.

Policy and Workplace Decisions Ahead

The next developments will depend on how healthcare employers adopt AI tools, what safeguards they establish for workers and patients, and whether lawmakers advance rules on employment uses of the technology. The California bill Khan mentioned was still under consideration according to the report; no next vote or timetable was provided.

Further reporting and research would be needed to determine how widespread the cited workplace practices are and whether they produce measurable changes in skills, wages, privacy or patient relationships. Khan’s remarks leave the central issue unresolved: how to use AI’s potential benefits while giving healthcare workers a meaningful role in decisions that affect their work.

Key Questions

What did Mishal Khan warn about?

Khan said AI may harm healthcare workers through de-skilling, monitoring, reduced autonomy and pressure on wages, even if it does not eliminate their jobs.

What happened to the medical residents in Uganda?

Hernandez-Boussard said a group of residents helping at a hospital cardiac ward faced a power outage and had no AI available to read EKGs. She said they lacked confidence synthesizing the information to make recommendations. The report did not give further details about the case or its outcome.

What worker-related laws did Khan cite?

Khan said Illinois and Vermont passed bills prohibiting the replacement of mental health workers with AI. She also cited a California law barring companies from advertising an AI chatbot as a health professional and a separate California bill under consideration about requiring workers to train AI replacements.

Is there evidence in the report that AI has already caused widespread harm in healthcare?

No. The report describes researchers’ concerns and one example involving medical residents, but it does not quantify how common the harms are or establish their overall effects across healthcare workplaces.

Source: rss