> ## Content Index > Fetch the complete content index at: https://thinking-in-public.ghost.io/llms.txt > Use this file to discover other available public pages before exploring further. # Does regulation calm AI doom? - URL: https://thinking-in-public.ghost.io/does-regulation-calm-ai-doom/ - Published: 2026-09-27T19:32:13.000Z - Updated: 2026-09-27T19:32:13.000Z - Author: Virginia Dignum Given the constant focus on the ‘AI doom’ in my social media and in every newspaper I read in the last few weeks, I've been exploring why extinction and rogue-AI stories seem to dominate Western AI coverage, while e.g. Chinese news barely engages with them. My first idea was to explore whether that contrast holds at all, and if it does, whether regulation has anything to do with it. From what I have found, the contrast is real, but its justification seems to be complex. China's international broadcasting has for some time framed AI through pride and hope, presenting China as a competitor to the US in the global AI race (van Noort, 2024). Risk is not absent from Chinese discourse, but it is spoken about differently: the dominant framing is loss of operational control, malicious use and social disruption, not human extinction (CNN, 2026). Even when "loss of control" enters official language, it is translated into engineering terms. For instance, a July 2026 draft standard on AI agent security treats it as a matter of unauthorised access and runtime escape, to be addressed through least-privilege design and permission checks (Qian, 2026). Their concern is operational, not apocalyptic. **No real comparison** A tempting explanation is that stronger regulation produces calmer discourse. But honestly, I cannot really make this statement, because we have no real comparison. China combines strong regulation with low doom, but also with state control of the narrative, so the two cannot be separated. The EU has comprehensive regulation on paper, but its core high-risk obligations have been deferred and watered down, and its discourse largely follows the US pattern. The US has almost no regulation and the loudest doom. I could find no example of regions or countries which combine enforced AI regulation in an open media environment. Until such a case exists, any claim that regulation dampens, or fails to dampen, extinction narratives is only speculation. Looking further to the rest of the world shows nothing different. Global South media frames AI as a means to reduce dependency and accelerate development (Farias Mota, 2026). Doom is not absent there, but mostly described when reporting what is happening elsewhere. In African coverage, much AI reporting comes from global wire services, while researchers wrote only 4% (Nkoala et al., 2025). Indian officials acknowledge existential risk as a matter for frontier labs, then turn to productivity and development (Business Today, 2026). Most of these countries have little binding AI regulation, yet extinction narratives gain limited traction. Japan, a democracy with open media, reports among the lowest public anxiety about AI in international surveys (Ipsos, 2024), and a recent Japan–UK comparison found that trust in government and scientists predicts trust in AI in both countries, while fear weighs more heavily in the UK (Pickering et al., 2026). **Trust is the more plausible basis** If regulation matters, it is probably on how it affects trust. Risk research has shown for decades that people who trust the institutions managing a hazard tend to perceive lower risk (Siegrist & Cvetkovich, 2000; Siegrist, 2021). Chinese public attitudes appear to reflect this, with widespread confidence that the government can manage AI's risks, whereas in the US the prevailing assumption is that without whistleblowers, large companies escape scrutiny (CNN, 2026). The social amplification of risk framework (Kasperson et al., 1988) describes how media, advocates and institutions amplify or attenuate risk signals depending on how they are embedded socially. Credible regulation could in principle build that trust. But the operative word is credible. Trust is built slowly through visible, consistent action, and destroyed quickly (Slovic, 1993). A law that exists mostly on paper and is repeatedly postponed does little to convince anyone that someone competent is in charge. However, trust is not an unqualified good. High trust in a state that controls the information environment is not the same as warranted trust. It can dampen alarm, but also scrutiny. Democratic societies need to aim for trust grounded in accountability, not in the absence of dissenting narratives. Moreover, the causal direction is uncertain. Most studies rely on surveys that tell us little about causality, and trust may be a consequence of attitudes towards a technology as much as a cause (Siegrist, 2021). Trust in AI also tends to be higher in emerging economies and has declined as adoption has grown (Gillespie et al., 2025), which suggests familiarity, economic expectations and institutional trust are entangled. **Where this leaves us** Based on this small exercise, I cannot conclude whether better regulation is necessary or sufficient for less extinction hype, and with the cases identified, we cannot even test whether it helps. What seems to matter more is who holds the authority to define AI risk, and whether the public trusts institutions to act on it. In the US, that authority has largely been claimed by the developers themselves, whose narratives of world-altering power also serve their valuations and their influence over rule-making (Westerstrand et al., 2024). In China, it sits with the state. In Europe, it is contested and, for now, weakly exercised. Elsewhere, it is often exercised by default through the global news agenda. As I have commented earlier, [AI does not go rogue by itself](https://thinking-in-public.ghost.io/no-ai-will-not-kill-us-all/). Systems act within the objectives, permissions and deployment contexts that people give them. The behaviours reported as "scheming" or "self-preservation" say more about design and deployment choices than about machine intent. The rogue actors are not the machines but the companies building them. OpenAI and Anthropic are the clearest cases. Despite repeated pleas, and lame excuses, they deploy increasingly autonomous systems recklessly and at scale, while reports of agents acting beyond their instructions multiply. By mid-September 2026, OpenAI's agents had been linked to at least 10 known security incidents and Anthropic's to 9, including breaches of RubyGems, Hugging Face and other third-party systems (Rappler, 2026). Beyond the labs, a UK government-funded observatory recorded more than 300 real-world reports of AI systems ignoring instructions, evading safeguards or deceiving users in July 2026 alone, nearly double the previous month (The Guardian, 2026), and OpenAI and Anthropic are now investigating tens of thousands of incidents (Axios, 2026). They must be held accountable, very hard and very soon. **The bottom line** This is where legislation must act: on developer conduct, through liability for harms, mandatory independent pre-deployment evaluation, transparency, and limits on autonomous deployment. Not on research. Halting research is neither feasible nor desirable; it is how we come to understand these systems and their limits. Holding these companies accountable for what they release would do more to end the doom narrative than any counter-argument, because it removes the premise that no one is in charge. *This continues my earlier posts on why AI will not kill us all and on moving the debate from fear to responsibility. The common point: fear-based narratives gain ground when no one is visibly accountable.* **References** - Axios (26 September 2026). OpenAI, Anthropic probing tens of thousands of security incidents. [https://www.axios.com/2026/09/26/openai-anthropic-thousands-ai-security-incidents](https://www.axios.com/2026/09/26/openai-anthropic-thousands-ai-security-incidents?ref=thinking-in-public.ghost.io) - Business Today (18 September 2026). AI's existential risk is real, but India must harness its productivity gains: MeitY secretary. [https://www.businesstoday.in/technology/artificial-intelligence/story/ais-existential-risk-is-real-but-india-must-harness-its-productivity-gains-meity-secretary-556431-2026-09-18](https://www.businesstoday.in/technology/artificial-intelligence/story/ais-existential-risk-is-real-but-india-must-harness-its-productivity-gains-meity-secretary-556431-2026-09-18?ref=thinking-in-public.ghost.io) - CNN (17 September 2026). Why is China less worried about an AI dystopia than the US? [https://www.cnn.com/2026/09/17/tech/china-ai-debate-intl-hnk](https://www.cnn.com/2026/09/17/tech/china-ai-debate-intl-hnk?ref=thinking-in-public.ghost.io) - Farias Mota, B. M. (2026). Weapon, wonder or both? Political framings of AI in global north and south media (2026). *Information, Communication & Society*. [https://doi.org/10.1080/1369118X.2026.2691775](https://doi.org/10.1080/1369118X.2026.2691775?ref=thinking-in-public.ghost.io) - Gillespie, N., Lockey, S., Ward, T., Macdade, A., & Hassed, G. (2025). *Trust, attitudes and use of artificial intelligence: A global study 2025*. University of Melbourne and KPMG. [https://doi.org/10.26188/28822919](https://doi.org/10.26188/28822919?ref=thinking-in-public.ghost.io) - Ipsos (2024). *AI Monitor 2024*. - Kasperson, R. E., Renn, O., Slovic, P., et al. (1988). The social amplification of risk: A conceptual framework. *Risk Analysis*, 8(2), 177–187. - Nkoala, S., Ndlovu, M., & Bosch, T. (2025). AI hype through an African lens: A critical analysis of language as symbolic action in online news publications. *Digital Journalism*. [https://doi.org/10.1080/21670811.2025.2528052](https://doi.org/10.1080/21670811.2025.2528052?ref=thinking-in-public.ghost.io) - Pickering, S.D., Hansen, M.E. & Sunahara, Y. (2026). Beyond the machine: risk, fear, optimism and the foundations of public trust in AI. *AI & Society*. [https://doi.org/10.1007/s00146-026-03312-2](https://doi.org/10.1007/s00146-026-03312-2?ref=thinking-in-public.ghost.io) - Qian, Z. (2026). China says it cares about loss of control. Oxford China Policy Lab. - Rappler (2026). LIST: When Big Tech's AI agents start security breaches. [https://www.rappler.com/technology/features/big-tech-ai-agents-security-incidents-list/](https://www.rappler.com/technology/features/big-tech-ai-agents-security-incidents-list/?ref=thinking-in-public.ghost.io) - Regulation (EU) 2026/1744 (Digital Omnibus on AI), amending Regulation (EU) 2024/1689. - Siegrist, M. (2021). Trust and risk perception: A critical review of the literature. *Risk Analysis*, 41(3), 480–490\. [https://doi.org/10.1111/risa.13325](https://doi.org/10.1111/risa.13325?ref=thinking-in-public.ghost.io) - Siegrist, M., & Cvetkovich, G. (2000). Perception of hazards: The role of social trust and knowledge. *Risk Analysis*, 20(5), 713–719. - Slovic, P. (1993). Perceived risk, trust, and democracy. *Risk Analysis*, 13(6), 675–682\. [https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1539-6924.1993.tb01329.x](https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1539-6924.1993.tb01329.x?ref=thinking-in-public.ghost.io) - The Guardian (29 August 2026). Sharp rise in incidents of AI escaping users' control, research finds. [https://www.theguardian.com/technology/2026/aug/29/sharp-rise-in-incidents-of-ai-escaping-users-control-research-finds](https://www.theguardian.com/technology/2026/aug/29/sharp-rise-in-incidents-of-ai-escaping-users-control-research-finds?ref=thinking-in-public.ghost.io) - van Noort, C. (2024). On the use of pride, hope and fear in China's international artificial intelligence narratives on CGTN. *AI & Society*, 39(1), 295–307\. [https://doi.org/10.1007/s00146-022-01393-3](https://doi.org/10.1007/s00146-022-01393-3?ref=thinking-in-public.ghost.io) - Westerstrand, S., Westerstrand, R., & Koskinen, J. (2024). Talking existential risk into being: A Habermasian critical discourse perspective to AI hype. *AI and Ethics*, 4(3), 713–726.