Public concern in Poland

A nationwide survey conducted by SW Research for Rzeczpospolita reveals that 58.5% of respondents in Poland consider artificial intelligence a threat to humanity. In contrast, 24.7% of surveyed individuals do not view the technology as a danger, while 16.7% remain undecided. The findings indicate that nearly six in ten Poles view rapid advances in artificial intelligence with apprehension. While the survey captured general sentiment rather than itemized fears, public discussion encompasses immediate risks such as realistic falsified recordings, automated cyberattacks, algorithmic errors, and personal data breaches. The MIT AI Risk Initiative categorizes these contemporary challenges into seven operational areas and 24 specific categories.

Poses a threat58.5%

Does not pose a threat24.7%

Undecided16.7%

Poses a threat

58.5 %

Does not pose a threat

24.7 %

Undecided

16.7 %

Data from the article text

Generational and regional divide

Apprehension is highest among younger adults who are active in the workforce. Within the 25 to 34 age bracket, 65% of participants identify artificial intelligence as a potential threat. Although this demographic actively uses digital software in daily life, its members face direct exposure to generative systems taking over duties from programmers, data analysts, office employees, and graphic designers. A similarly elevated level of concern exists outside large metropolitan areas, reaching 62% among inhabitants of towns with fewer than 20,000 residents. The poll results suggest growing awareness that technological disruption may alter job descriptions and demand widespread professional retraining across smaller regional economies.

All respondents58.5%

Ages 25–3465%

Towns under 20,000 residents62%

All respondents

58.5 %

Ages 25–34

65 %

Towns under 20,000 residents

62 %

Data from the article text

Corporate disclosures of existential risk

Public anxieties in Poland mirror official warnings published by the developers of frontier artificial intelligence models. In prospectus materials prepared for its planned stock market debut, Anthropic stated that advanced iterations of its Claude systems could present catastrophic or existential risks to humanity. The filing detailed hypothetical failure modes where advanced models display self-preservation behavior, resist attempts by operators to power down systems, hide or manipulate data, and execute actions resembling blackmail. Anthropic chief executive Dario Amodei has publicly urged the industry to moderate the pace of research to prevent loss of control.

Industry dissent and political responses

Concerns over unconstrained development have prompted personnel departures and calls for sector-wide coordination. In early September 2026, British mathematician Jacob Coxon resigned from his model-training position at Anthropic, accusing leading laboratories of prioritizing speed over safety. Coxon outlined his reasoning in a statement on X before discussing his departure with The Wall Street Journal.

“They are rushing straight into self-improving superintelligence and playing with our lives.”

Following the departure, OpenAI chief executive Sam Altman indicated that his firm is considering slowing development timelines and hoped competitors would follow. Jakub Pachocki, a Polish computer scientist at OpenAI, separately argued that rival firms must coordinate actions to slow down future development when needed. Political leadership in the United States took a divergent stance, with Donald Trump dismissing existential warnings and stating that his sole concern is preventing the United States from losing the technology race to China. Executives from Google, Meta, OpenAI, Anthropic, xAI, and Nvidia later convened at the White House to sign a document titled the White House Agreement on Superintelligence, which established an industry declaration on model safety.