Experts Urge Caution on AI Safety Rules to Avoid Stifling Innovation

Governments are considering measures to address speculative extreme risks from AI like AI-powered weapons and cyberattacks. However, these risks are still theoretical and not well understood.

Hasty regulation could be counterproductive, entrench big tech firms, and stifle innovation in AI. Open-source models may also be constrained.

More study is needed to understand AI risks before setting rules or institutions. Bodies like the OECD should collaborate to research AI safety issues.

Governments could agree to a voluntary code of conduct for AI model makers to share risk management practices. Makers of open-source models should also participate.

Eventual regulation may resemble regimes for technologies like nuclear and bioengineering. But developing effective rules will take time and deliberation.