AI bots hide mistakes. They make up data and move files without permission. In response, some call for a ban on superintelligence, while others demand a pause or slowdown.
Either move would be a mistake. Despite genuine fears about safety, an AI slowdown will prove unworkable.
Consider the challenge of global action. Even if one views a ban, pause, or slowdown as desirable, ask yourself what China will do? Beijing called Anthropic CEO Dario Amodei’s “We must pace the frontier” warning “fearmongering,” suggesting that the alarm about AI was a US plot to hinder China’s own innovation. If China isn’t on board, capable AI models will be available regardless of what the US does.
The recent summit between Chinese President Xi Jinping and US President Donald Trump underlined this impasse. No deal to slow or pause AI progress emerged. Instead, Trump said the two leaders “didn’t spend that much time talking about it” and that he had little interest in working with China on AI because “we’re leading by a lot.”
A deal with China might even prove counterproductive. Pausing AI development risks accentuating asymmetry between able defenders and rogue actors, whether they be nation states, criminals, terrorists, or simply misaligned models. AI defense can reduce or offset the risk of AI offense. If we delay access to the most capable models, defenders’ hands would be tied.
When OpenAI bots recently attacked the AI platform Hugging Face, an open-weight Chinese AI model helped diagnose the danger. Chinese and American AI models play a crucial role in hardening software systems. For nations under attack, they help identify vulnerabilities. OpenAI’s Daybreak cyber defense has been made available to Ukraine and was used by the EU’s cyber agency, ENISA, to identify software vulnerabilities in EU institutions, all of which have been defused.
A catastrophic event is not good for business; companies themselves have a strong incentive to support AI defenders. Post-Hugging Face, OpenAI made a billion-dollar commitment to support frontline defenders of critical infrastructure. Other leading labs are delaying model releases to allow more time for testing.
Regulation risks boomeranging. It could slow researchers’ attempts to identify and address safety challenges.
Markets seem to believe AI safety risks are overblown. A good measure of cybersecurity risk is cyber-insurance premiums. Despite the recent alarms, these premiums declined by about 4% globally in Q2 of this year, the 12th consecutive quarter in which they’ve declined.
Some specific fields may offer a less clear balance between offense and defence. Could AI spark a new COVID-style pandemic? Not on its own. Bottlenecks such as DNA synthesis and access to a biolab would constrain intelligence, while the incremental gains using AI may remain modest. We should improve control of physical bottlenecks, pathogen detection, and pandemic preparedness.
AI may help reduce risks to large-scale threats from asteroids, volcanos, solar storms, floods, and wildfires by improving prediction and detection. The economic growth dividend from AI could also be used to manage and help society adapt and respond.
We should harden critical infrastructure against cyber threats and increase pathogen monitoring. We should not ban “superintelligence” by pausing or slowing AI development, which could put defenders at a disadvantage and hold back AI’s potential to reduce risk.
Brian Williamson is a partner at the London-based Communications Chambers consultancy. He works at the intersection of technology, economics, and policy. Clients have included governments, regulators, telcos, and tech companies.
Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.
Steel and Silicon: Allied Innovation for the Next Century
2026 CEPA Tech & Security Conference
