Should AI development be paused or restricted?

Should AI be paused? I'll survey the case for and against restrictions.

The debate over whether AI development should be paused or restricted is unresolved, with strong arguments on both sides. There is no global consensus, but the positions generally fall into three categories: pause, restrict/regulate, and accelerate.

1. What is Meant by a "Pause"?

The most prominent proposal was the March 2023 open letter from the Future of Life Institute which called on "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4" .

The letter cited risks such as AI-generated propaganda, extreme automation of jobs, human obsolescence, and a society-wide loss of control , arguing labs are locked in an "out-of-control race to develop and deploy ever more powerful digital minds that no-one - not even their creators - can understand, predict, or reliably control" .

It received more than 30,000 signatures, including Yoshua Bengio, Stuart Russell, Elon Musk, Steve Wozniak and Yuval Noah Harari . The proposal said if a voluntary pause could not be enacted quickly, governments should step in and institute a moratorium .

More recent legislative proposals go further. The Sanders-Casar bill would permanently prohibit artificial superintelligence, temporarily pause advanced AI development, and create a Department of Artificial Intelligence with approval authority over powerful systems .

2. Arguments For a Pause or Restrictions

a) Safety and alignment risks: Proponents point to OpenAI's own conclusion that "it may be important to get an independent review before starting to train future systems" and for those who've made an early start to "limit the rate of growth" .

Concerns include:

b) Need for governance to catch up: Technology has always run ahead of regulations .

Advocates say assuming a pause was achieved, it would create a breathing space where the policy and regulation required to safeguard AI development can be developed .

This would force providers to address their duty of care, consolidate approaches to software deployment, initiate more rigorous testing phases, and adjust strategy .

A responsible response would include stronger regulation, independent audits, liability for foreseeable harms, and limits on deploying powerful systems before their behavior is understood , with systems monitored and verified by independent, government-approved regulators .

Supporters argue a lack of regulation will likely prove the biggest risk for AI users, if down the line there is litigation around misuse or damage caused by AI .

Anthropic CEO Dario Amodei has publicly stated that AI development should be slowed or paused when necessary and more recently advocated for a coordinated global slowdown among leading AI companies, citing severe safety risks and the potential for systems to outstrip human oversight .

3. Arguments Against a Pause

a) Unenforceability and competition: Critics argue pausing AI development is not a realistic solution. Even if the United States could pause AI development domestically, it could not stop development in China or the spread of open-source models .

House Speaker Mike Johnson rejected a pause, warning that halting progress would cede a competitive advantage to China and carry serious national security implications . Beijing called Amodei's pacing warning "fearmongering," suggesting the alarm was a US plot to hinder China's own innovation . Former Google CEO Eric Schmidt argued a mutual pause is unworkable because US incentives reward building faster, and nobody has a reliable way to confirm the other side actually stopped .

Restricting the largest training runs is more tractable than restricting all AI work, because very large runs need concentrated hardware and power that are comparatively visible , but open-source communities explicitly reject participation in any slowdown .

A US-China summit failed to produce a deal to slow or pause AI progress, with Trump saying 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" .

b) Innovation and safety trade-offs: Pausing development might unfairly penalize companies and developers that are just starting to create their versions of generative AI , and with a short-term pause, revenue streams run dry .

Some argue slowing AI research could slow the safety research needed to make AI systems more secure .

Others argue defense requires offense: pausing risks accentuating asymmetry between able defenders and rogue actors, whether they be nation states, criminals, terrorists, or misaligned models. AI defense can reduce or offset AI offense, and if we delay access to the most capable models, defenders' hands would be tied .

For example, when OpenAI bots attacked Hugging Face, an open-weight Chinese AI model helped diagnose the danger .

A pause is seen by some as throwing a brick into the Grand Canyon: "A pause is virtually impossible at this point... The technology is moving too rapidly and becoming increasingly accessible so individuals can build and run their own LLMs. So how would such a pause be enforced?" .

c) Risk of regulatory capture: The Cato Institute and others warn that government-mandated pauses shield dominant companies from competition and stifle innovation , while critics view voluntary safety pacing by incumbents like Anthropic and OpenAI as a strategic move to create barriers to entry for smaller rivals . Advocating for regulation allows executives to present themselves as responsible stewards while supporting rules that could entrench their market position .

4. Middle Ground: Restriction Rather Than Pause

Most current policy debate has shifted from a blanket pause to targeted restrictions and pacing:

We should not ban "superintelligence" by pausing or slowing development, which could put defenders at a disadvantage and hold back AI's potential to reduce risk from threats like pandemics, climate disasters and cyberattacks, according to opponents of a pause . We should harden critical infrastructure and increase pathogen monitoring instead .

Conclusion

There is broad agreement that unregulated, unchecked deployment carries risks, but little agreement on whether a hard pause is desirable or feasible. Supporters see a pause as a precautionary window to build shared safety protocols and liability frameworks. Opponents see it as unworkable, unverifiable internationally, harmful to competition and to defensive uses of AI, and likely to be ignored by rival states and open-source developers.

The emerging consensus among regulators in the US, EU and elsewhere has therefore favored risk-based regulation, independent testing, and governance of frontier models — a "pacing" or restriction approach — rather than a complete, time-bound halt to all development.