Archive · 16 Sep 2026
AI Leaders Clash Over Pace, Safety, and Regulation
Several leaders in artificial intelligence discussed how to make AI safer.
Jensen Huang said safety should be treated like an engineering problem.
He said companies should build testing environments and check their products carefully.
If a company is not confident that an AI product is safe, Huang said it should wait before releasing it.
He also said companies can move quickly while still making safe products.
Sam Altman said companies should act responsibly even when competitors or countries are racing to develop AI.
Dario Amodei said companies should look honestly at their own safety records instead of only criticizing competitors.
The leaders agreed that safety matters, but they differed over how much the industry should slow down and whether new regulations are needed.
NVIDIA CEO Jensen Huang said companies should test AI systems and pause releases when they lack confidence in product safety.
Huang argued that rapid development and safety can coexist, and said new laws and regulations are unnecessary.
OpenAI CEO Sam Altman said companies should act responsibly regardless of competitors, commercial pressure, or international competition.
Anthropic CEO Dario Amodei urged companies to examine their own safety records, improve practices, invest more in safety, and increase transparency.
The executives discussed AI safety at Salesforce’s Dreamforce conference in San Francisco on September 16.
- Who
- NVIDIA CEO Jensen Huang, OpenAI CEO Sam Altman, and Anthropic CEO Dario Amodei.
- What
- The executives debated AI safety, development speed, testing, transparency, industry coordination, and regulation.
- When
- September 16.
- Where
- Salesforce’s Dreamforce conference in San Francisco, California, United States.
- Why
- They were addressing concerns that rapid competition among companies and countries could encourage unsafe AI development.
Rapid Development With Self-Regulation
Deliberate Pacing and Shared Safeguards
Development speed
Rapid Development With Self-Regulation
Jensen Huang said companies can pursue fast AI progress while building safe products, describing a necessary choice between speed and safety as false.
Deliberate Pacing and Shared Safeguards
Dario Amodei has called for the industry to slow the pace of frontier-model improvement, and Sam Altman has backed pacing the frontier without stopping development.
Need for new rules
Rapid Development With Self-Regulation
Huang said new laws and regulations are not needed, arguing that companies should test products and decide when they are safe enough to release.
Deliberate Pacing and Shared Safeguards
Amodei has proposed independent safety evaluators within AI companies, common safety standards among leading democracies, and broader international cooperation.
Responsibility during competition
Rapid Development With Self-Regulation
Huang emphasized company-level engineering, testing, and market decisions as the main safeguards.
Deliberate Pacing and Shared Safeguards
Altman said safety should not depend on what competitors or countries do, while Amodei urged companies to inspect their own shortcomings and increase transparency rather than merely criticize rivals.
- Main debate
- Whether AI companies can maintain rapid development while ensuring product safety.
- Jensen Huang’s position
- Companies should test AI systems and pause releases when they are not confident about safety.
- Regulation
- Huang said new laws and regulations are unnecessary and that companies should regulate themselves through engineering, testing, and market decisions.
- Sam Altman’s position
- Companies should commit to responsible AI without making safety dependent on competitors’ actions or commercial pressure.
- Dario Amodei’s position
- Companies should review their own safety records, improve their practices, recommit to transparency, and invest more in safety.
- Different pacing views
- Amodei has called for slowing the pace of frontier-model development, while Huang said speed and safety are not mutually exclusive.
- Event host
- Salesforce CEO and Chairman Marc Benioff hosted the event.
