The Price of Conscience: Why Voicing AI Ethics Concerns Makes You an Industry Pariah

Martyn Berlin chose his side years ago. He knows the technology. He grasps its mechanics and its consequences. So he stands vehemently against the current form of AI. No positive outcome, he argues, justifies the damage already inflicted and still unfolding. The environmental toll, exploited labor in data labeling, theft from creators, erosion of critical thinking, concentration of power, flood of disinformation, degradation of the internet and the gutting of entire professions leave him convinced. “Just one more model” will not deliver genuine intelligence. For Berlin, the term AI itself serves mostly as marketing.

That position isolates him. In technology circles and beyond. Friends drift away. Communities close off. A theater group once snapped a photo and turned it into a band poster using ChatGPT without asking. The result left him sickened. Another acquaintance casually let Siri tap ChatGPT for medical advice on medication duration and accepted the output without question. Berlin no longer wants to spend time with him. Presentations that criticize AI flaws yet demonstrate them live with Copilot prompt him to walk out. The constant barrage exhausts him. Even Wikipedia suffers as users query models trained on its content, accept hallucinations, and never bother to edit or contribute back.

Stomach-turning. That’s how Berlin describes seeing respected people endorse or deploy the tools openly. He has reached the point of severing ties with groups that promote its use. The linked essay he references, from smallsheds.garden, captures a similar fatigue with widespread acceptance. And he does not stand alone in feeling the strain. Recent reporting shows his experience reflects a broader pattern.

Across Silicon Valley and beyond, those who raise alarms about training on copyrighted material, biased outputs, job displacement or unchecked power face professional exile. Ed Newton-Rex quit his role as vice president of audio at Stability AI in late 2023. He could not reconcile the company’s practices with a society that depends on copyright to sustain creative work. He founded Fairly Trained, a nonprofit that certifies models built on properly licensed data. (The Atlantic)

Timnit Gebru’s dismissal from Google in 2020 followed her co-authorship of a paper highlighting how large language models could amplify biases against marginalized groups. The episode triggered an open letter signed by thousands and congressional scrutiny. Suchir Balaji left OpenAI in 2024 after concluding its use of copyrighted works violated the law. He planned to testify in related lawsuits before his death by suicide later that year. Leopold Aschenbrenner, a former OpenAI safety researcher, was fired after flagging what he saw as inadequate security around critical model information. (Untold Magazine)

But the personal cost extends past formal whistleblowers. Engineers, researchers and even mid-level managers who question deployment timelines, data practices or societal impacts report being sidelined, passed over for promotion or quietly edged out of key projects. The industry rewards acceleration. Doubters slow it down. Or so the thinking goes.

Public sentiment has turned sharply. Only 26 percent of voters hold positive views of AI while 46 percent feel negative, according to one survey. Just 10 percent of Americans express more excitement than concern. Data centers have become flashpoints. Maine attempted the first statewide moratorium, though the governor vetoed it. A record number of projects were canceled in early 2026 after local opposition. In Indianapolis, someone fired 13 rounds at a councilman’s home with a note reading “NO DATA CENTERS.” Another man hurled a Molotov cocktail at Sam Altman’s residence before threatening OpenAI’s headquarters. Social media posts celebrating the act gathered thousands of likes. (The Atlantic)

Steve Bannon and Bernie Sanders, unlikely allies, both frame AI as a threat to ordinary workers. Sanders wrote that AI oligarchs seek to replace workers themselves, not merely specific jobs. Bannon has accused Silicon Valley of ignoring the little guy. Senator Mark Warner voiced enormous concern that populism from left and right could stifle innovation. Nathaniel Persily, Stanford law professor, observed that many Americans doubt they will number among the winners, drawing on two decades of technological history. Senator Josh Hawley questioned whether the coming riches will benefit children, parents or the American worker. (The Atlantic)

Gary Marcus has watched the backlash build for years. He warned of many current problems in his 2023 Senate testimony. OpenAI itself flagged similar risks in a 2019 report. Generative AI, outside narrow uses like coding assistance or brainstorming, has delivered a net negative, he contends. It undermines education, fuels disinformation and deepfakes, enables bias and crime, widens economic gaps, and drives construction of power-hungry data centers that damage the environment. “It’s no wonder many people have had enough,” Marcus wrote recently. He predicts anti-AI sentiment will shape the 2028 presidential race. (Gary Marcus on Substack)

Executives have begun to adjust their language. Altman once stated plainly that jobs would disappear. Now he calls job doomerism likely wrong in the long term. Venture firms publish essays dismissing apocalypse narratives. Yet the disconnect persists. Those earning over $200,000 feel most optimistic about AI in daily life. Everyone else sees concentrated wealth, disrupted communities and uncertain futures. (The Atlantic)

Berlin never sought to become a symbol. He simply refuses to set aside his ethics for convenience or career. That choice carries weight. He loses friends. Influence slips. Tears come when entire chapters of life close. Some must use these tools to keep their jobs. He extends sympathy there. Voluntary embrace after understanding the harms draws sharper judgment. Promotion of the technology, however subtle, prompts him to leave groups or avoid individuals entirely.

His stance is not unique. It echoes across research labs, creative fields, newsrooms and policy circles. The Council on Foreign Relations documented rising unpopularity and political pressure for regulation. Over 1,200 AI-related bills appeared in state legislatures in 2025 alone. Protests target data centers. Unions and artists push back against training data practices. Students boo AI-focused commencement speeches. (Council on Foreign Relations)

And yet the machine rolls forward. Valuations climb toward trillions. Models grow more capable. Pressure to deploy intensifies. Those inside who voice caution risk being labeled obstructive, naive or worse. Outcasts. The term fits. It carries real emotional and professional consequences. Berlin describes the toll as unrelenting. No week off from the advertisements, the casual adoption, the assumption that resistance equals backwardness.

Critics inside the industry sometimes face retaliation cloaked in nondisclosure agreements or equity clawbacks. One former OpenAI employee forfeited roughly $2 million in vested shares rather than sign a non-disparagement clause. Others stay silent to protect their careers. The pattern repeats. Question the extraction of creative work without compensation and find yourself on the wrong side of legal strategy. Highlight safety shortfalls and watch projects accelerate anyway. Point to societal risks and hear accusations of fearing progress.

Recent months have only sharpened the divide. Violence against infrastructure and executives, though still rare, signals deeper anger. Political operatives on both sides test populist messaging that casts AI as a tool of unchecked corporate power. Local officials seek to hide their addresses after threats. The structural conditions for broader unrest exist, researchers warn. History offers parallels in the Industrial Revolution, when machines became targets amid hardship and inequality.

Berlin’s essay captures a quieter, more intimate form of that conflict. Not protests or legislation. The daily choice to maintain principles at the expense of belonging. It sucks, he says plainly. The isolation. The judgment from others who view his position as extreme. The effort required to keep explaining why the trade-offs feel unacceptable. He does not expect others to adopt his morals. He simply will not abandon them to fit in.

That quiet resolve appears more frequently now. In private conversations among researchers. In resignations that never make headlines. In the growing number of professionals who refuse to list certain tools on their resumes or who steer client work away from generative systems. The industry notices. Some respond with better messaging. Others double down on acceleration. Few seem prepared to address the root discomfort: many who understand the systems best find their deployment morally untenable.

Public polls, violent incidents, legislative activity and personal testimonies all point the same direction. The backlash has arrived. Its form will shift as capabilities advance and effects spread. For those already living with the consequences of their ethical positions, the future looks lonely. They knew the technology. They weighed the impacts. They chose conscience anyway. The outcast label followed. And it still sucks.