Are the people building AI exposed to particular psychosocial risks?

A particular combination of already known risks

A recent Business Insider investigation among therapists supporting Silicon Valley professionals highlights several unusual situations. Engineers question whether their expertise will remain valuable. Some workers feel uneasy about the consequences of technologies they helped develop. Others remain in highly demanding organisations because a considerable share of their future compensation still depends on staying.

These accounts concern technology workers more broadly and do not isolate employees of AI laboratories. They nevertheless provide an interesting signal for those working closest to the most advanced systems.

The issue is probably not a set of new psychosocial risks unique to AI. It is a particular combination of already known risks, produced jointly by technological acceleration and the organisations surrounding it.

Caution: the accounts concern the technology sector broadly. They identify a hypothesis for study, not a phenomenon already demonstrated among AI workers.

Building the tool that may devalue one’s own expertise

Some AI workers directly help improve systems that may automate a growing share of their own activity.

Unlike many other professions, they do not necessarily discover this progress through the press. They may directly observe what models could not do a few months earlier and then take part in reducing the remaining limitations.

Business Insider reports that several senior engineers seeing one therapist no longer imagine doing the same job in ten years. This uncertainty appears to concern something deeper than employment security alone. One therapist reports the following question among some patients:

“I don't know who I am if my work gets automated.”

The development also creates a paradox. AI can greatly increase an engineer’s capabilities today while weakening the future value of that expertise.

The same professional may therefore feel far more capable now and far less certain about what will constitute their professional value tomorrow.

An immediate increase in the sense of power can coexist with growing uncertainty about the future value of one’s skills.

This corresponds to the job insecurity described among the major psychosocial risk factors. But the perceived threat no longer concerns only the current job. It may also affect the basis of recognition and professional identity.

Highly autonomous in the work, far less so over its direction

Highly qualified AI workers may have considerable technical autonomy. They choose methods, conduct complex research and make daily decisions requiring rare expertise.

Yet this autonomy does not mean they have the same power over the overall direction of the systems they develop.

An engineer may decide how to improve a model without deciding its release schedule. A safety specialist may identify a risk without having the power to prevent deployment alone. A team may master the technical solution while having much less control over the organisation’s economic or strategic objectives.

We must therefore distinguish operational autonomy from decision-making power. This separation may become especially important when the anticipated consequences of the work are considered major.

Conflicts of values and ‘complicity grief’

Psychotherapist Annie Wright uses the expression ‘complicity grief’ to describe the unease of professionals who help build technologies while fearing some of their consequences.

She gives the examples of an engineer who trained a model that later displaces some jobs, a product manager who deployed a feature used to reduce staffing, and a founder developing an agent capable of making an entire profession seem ordinary.

The expression is neither a medical diagnosis nor a recognised psychosocial-risk category. But the experience described directly evokes conflicts of values: situations in which a person believes their work contributes to consequences that are difficult to reconcile with their convictions.

Wright also uses the expression ‘moral injury’ for an employee in conflict with what she helps produce. Moral distress and moral injury are the subject of growing scientific literature in several professional settings, but have not been established specifically among AI workers.

The most relevant configuration may be less guilt itself than the combination of high felt responsibility and limited decision-making power. A worker may feel personally implicated in a technology’s consequences while having only partial influence over how it is eventually used.

When meaning makes limits harder to set

Some AI workers feel they are taking part in a historic transformation. This belief can be a considerable resource. It gives meaning to work, stimulates learning and may explain an exceptional level of commitment.

But meaning is not necessarily protective against excessive workload. Competition between companies, rapid progress and the need to remain continuously at the frontier of knowledge can encourage very intense working rhythms. The Business Insider investigation also describes environments marked by permanent availability and career paths in which long hours gradually become normal.

The risk is that passion does not compensate for intensity but instead helps make it acceptable.

Therapists describe another phenomenon after a long-pursued objective has been achieved. Following an initial public offering or access to considerable wealth, some workers do not feel the expected relief but rather emptiness or disorientation. Wright calls this the ‘arrival paradox’.

After years in which identity was built around work, professional progress and an objective to be reached, the sudden disappearance of that tension may itself become destabilising.

The issue is therefore not only working a great deal. It may also be that work has gradually occupied a central place in the definition of the self.

The meaning attached to work can be a resource while also making a persistently excessive workload seem more acceptable.

When exceptional remuneration reduces the freedom to leave

The ‘liquidity anxiety’ described in the article should also be examined from an occupational-health perspective. It is not financial insecurity in the usual sense.

Some workers may possess considerable wealth on paper in the form of shares, stock options or deferred compensation. But that wealth may only become accessible after particular deadlines or may depend directly on remaining with the company.

Wright summarises the situation particularly clearly:

“technically rich on paper and functionally stuck in real life.”

A worker may therefore appear extremely wealthy while believing it would be financially very costly to leave now. Business Insider refers in particular to vesting schedules that can make departure especially difficult.

Compensation becomes ambivalent. It rewards investment in the organisation while simultaneously increasing the cost of exit.

In a company where workload becomes excessive, or whose direction conflicts with a worker’s values, this compensation architecture may concretely reduce the capacity to leave.

Wealth becomes both a resource and a dependency.

The cross-cutting role of professional identity

One element recurs across several accounts in the investigation. The personal value of some professionals appears strongly tied to their job title, company prestige, level of compensation, speed of promotion or ability to remain among the best.

This strong identification with work is of course not specific to artificial intelligence. But technological acceleration creates a distinctive configuration when the expertise underlying that identity itself seems likely to lose value rapidly.

The same work may then produce recognition, wealth and a sense of power while also creating a threat of obsolescence.

Professional identity links the value of expertise, prestige, remuneration, meaning and the ability to contemplate leaving.

A work-organisation issue before an individual vulnerability

This is probably the most important point in the investigation.

Expressions such as ‘complicity grief’ and ‘liquidity anxiety’ could easily suggest that new psychological vulnerabilities specific to Silicon Valley workers are emerging.

Yet Annie Wright believes the industry itself ‘bears substantial structural responsibility’.

She points in particular to permanent availability, deferred compensation that makes leaving financially costly, and organisations built around a continuous pursuit of growth and performance.

Behind new expressions, we therefore mainly find much more familiar organisational determinants: high intensity, conflicts of values, job insecurity, limited control over some decisions, economic dependence and temporal boundaries that are difficult to maintain.

This distinction is essential for prevention. The goal is not merely to help individuals tolerate technological change more effectively. It is also to examine working conditions and incentive structures that contribute to producing these situations.

An unusual combination, not yet an established phenomenon

A particular configuration is beginning to emerge.

An AI worker may have exceptional expertise while doubting its durability. They may have great technical autonomy while exercising little control over the general direction of the technology. They may find enormous meaning in their work while finding it more difficult to set limits. They may feel significant responsibility while holding limited decision-making power. Finally, they may receive considerable compensation while facing a high cost if they decide to leave.

The paradox may not be that AI workers face new psychosocial risks. It may be that they combine exceptional professional resources with—sometimes through those very resources—unusual forms of dependence, uncertainty and conflict with their own work.

None of these mechanisms is unique to artificial intelligence. What may be more specific is their combination among people working closest to systems whose capabilities are advancing rapidly and whose consequences they themselves anticipate.

At this stage, this remains a hypothesis. Available data mainly concern technology workers, therapists’ accounts and non-representative investigations. They allow neither estimation of how frequent these situations are among AI workers nor a causal link between developing AI and mental-health disorders.

They nevertheless raise a question that now deserves study: are the people building advanced AI systems exposed to a particular configuration of psychosocial risks produced by technological acceleration, work organisation and the threat of self-obsolescence?

What this analysis can—and cannot—establish

• Established framework — Work intensity, job insecurity, conflicts of values, autonomy and recognition are established dimensions of psychosocial-risk analysis.
• Emerging signal — The reported accounts describe self-obsolescence, ‘complicity grief’, the ‘arrival paradox’ and ‘liquidity anxiety’ among some technology workers.
• Hypothesis — The combination of technological acceleration, work organisation and threatened self-obsolescence may be particularly marked among those developing advanced AI systems.
• Limit — Journalistic investigation and clinical accounts cannot estimate prevalence or establish causation.

Sources and cautious reading

Business Insider, republished by AOL, August 2026: accounts of anxiety among Silicon Valley workers and their therapists, including questions about self-obsolescence, ‘complicity grief’ and ‘liquidity anxiety’. These are accounts and signals, not epidemiological measurements.

Gollac and Bodier, 2011, Measuring psychosocial risk factors at work: the framework used here to relate the accounts to work intensity, conflicts of values, autonomy, recognition and insecurity of the work situation.

Expressions drawn from personal accounts illuminate possible experiences. They do not replace established prevention categories or research data.