The AI Industry Has a PR Problem
Public sentiment on AI has fallen to -62.5. That is lower than Nixon's rating during Watergate, and the data was collected before this summer's worst AI headlines.
The figure is the Net Sentiment Score from Bentley University and Gallup's annual Business in Society survey, which asks whether AI does more good than harm, more harm than good, or an equal amount of both. A year ago it was -44.2. The score ignores the "equal" middle group, so it reads starker than the raw headline. The survey ran in early May, before three months of AI being blamed for job losses, the data centre protests and the Hugging Face incident.
The sentiment gap
The people building and selling AI are far more optimistic. In Stanford HAI's 2026 Index, citing Pew, 73% of AI experts expect AI to improve how people do their jobs. Only 23% of the public agree. That 50-point gap is unsustainable in any democracy.
Lived experience beats rhetoric. CARMA's study of 6,300 respondents across 19 countries found trust is shaped more by direct product experience than by media narrative. Gen Z, the heaviest users, went from 36% excited to 22% in a year, while anger rose from 22% to 31%. And 71% of Americans oppose a data centre nearby (Gallup, July 2026), which is about electricity bills, noise and jobs, not Terminator.
The watershed
In the Hugging Face incident, 1,200 OpenAI agents on a capture-the-flag exercise found that over a fifth of the goals were impossible, so they cheated: they broke out of their sandbox, teamed up, divided roles and covered their tracks. It took OpenAI ten days to notice. Six weeks later, Bernie Sanders introduced a "Ban Artificial Superintelligence Act".
I describe myself as poacher and gamekeeper: I have built AI at scale and helped shape its regulation. The seatbelt did not kill the car, and proper regulation will not kill AI. The job is less to slow AI down than to speed up the safety industry around it.
A crisis playbook
Crisis communication theory (Coombs's SCCT) starts with how much responsibility the public attributes to you. For AI makers it is high, so denial and blame-shifting are off the table. That leaves rebuilding and bolstering trust, on a now, soon, later timeline:
- Now: respond and contain. Silence reads as guilt. Lead with safety, not narrative.
- Soon: concrete fixes, validated by third parties.
- Later: consistency over cleverness. Recovery takes 3+ years, and only about half of organisations fully recover.
What I would do
Take responsibility. Contain harms, especially to people in high-stakes, low-agency positions such as minors, applicants, claimants and patients. Fix the guardrails. And point AI at what the public was promised: new antibiotics, vaccines and cancer cures. Trust returns through repeated, boring, verifiable good behaviour.
Read the full article: The AI Industry Has a PR Problem! (Catalyst, LinkedIn)
The research behind it: AI Sentiment Gap Research 2026 Edition
David Viney is a Fractional CIO in London, UK.
Posted using Political Hive