Employers Deploy Emotion‑Reading AI Across Workflows, Prompting Scientific and Legal Scrutiny

Companies are increasingly deploying software that claims to read employees' emotions during meetings, job interviews and customer interactions, a May 9, 2026 feature by Ellen Cushing reports. The tools are moving from pilot projects into routine corporate workflows, raising questions about scientific validity and real‑world consequences for workers who may be surveilled or evaluated by automated systems. Cushing describes hands‑on testing and a range of real deployments. In one experiment, vendor MorphCast analyzed her facial expressions during a meeting and returned labels such as "amused," "determined," "interested" and sometimes "impatient.

and platforms from Slack (Aware) and Microsoft Azure that offer sentiment analysis for internal communications. Hiring vendors such as Imentiv market emotional analysis for candidate screening. The feature preserves both vendor claims and scientific critiques. Many products rely on Paul Ekman's basic‑emotions framework, a model long disputed by researchers as overly simplistic. Neuroscientist Lisa Feldman Barrett is quoted emphasizing that facial and bodily movements do not have fixed emotional meanings but instead convey relational signals. Cushing also cites empirical work, including a study by Lauren Rhue, showing systems can misread emotions in racially biased ways-for example, labeling Black NBA players as angrier than white teammates in some cases.

Regulatory and market forces are complicating rollout. The European Union has banned emotion‑reading AI in the workplace under the AI Act, with narrow exceptions for medical and safety uses; Cushing notes that MorphCast responded by moving its headquarters from Florence to the San Francisco Bay Area. At the same time, an estimate cited in the piece projects the global market for affective‑computing products could triple to roughly $9 billion by 2030, suggesting commercial pressure to keep developing and selling these tools.

The story documents concrete harms and contested outcomes tied to workplace monitoring. A 2022 investigation found UnitedHealth social workers were downgraded for keyboard inactivity while they were communicating with patients, and the ACLU alleged that HireVue and its client Intuit denied a promotion to a deaf accessibility‑team employee based on automated "active listening" checks — claims both companies have disputed. Those examples illustrate how surveillance and automated assessments can have operational effects on employees' careers and evaluations.

For builders, deployers and regulators, the report highlights practical priorities: stronger validation of scientific assumptions, robust bias testing, legal clarity across jurisdictions, and ethical limits on workplace surveillance. Cushing closes by warning of a further risk beyond flawed tools — an environment in which employees feel compelled to perform emotions to satisfy automated monitors, with consequences that extend beyond individual technical choices.