Nvidia CEO Urges AI Safety via Engineering, Not New Regulation
Nvidia CEO Jensen Huang says AI safety should rely on engineering, not new laws.
Why it matters: Legal and regulatory professionals must assess the call for engineering-led AI safety against evolving demands for new AI governance frameworks and compliance standards.
- Nvidia CEO Jensen Huang told the G20 Innovation Ministerial on Sept. 2 that AI safety is best addressed through engineering, not legislation.
- Huang argued existing laws suffice since AI involves human-made hardware and software requiring no new legal frameworks.
- Anthropic CEO Dario Amodei warned AI-powered botnets could cause massive internet damage within 6 to 12 months, highlighting real-world AI risks.
- Nvidia and Palantir are limiting use of advanced AI models due to privacy and intellectual property concerns, revealing operational compliance challenges.
At the September 2 G20 Innovation Ministerial, Nvidia CEO Jensen Huang emphasized that AI safety is primarily an engineering problem, not a legal one. "Safety is an engineering problem, not a legal one. We’re developing software after all," Huang stated to TechCrunch.
Huang pointed out that AI systems consist of hardware and software components designed by humans and therefore can be managed under existing legal frameworks without creating new laws. He urged regulators to focus on tangible risks rather than "theoretical harms." This perspective challenges calls for stricter regulations and suggests market incentives will encourage companies to develop safe AI.
Contrasting Huang's views, Dario Amodei, CEO of Anthropic, cautioned legal and compliance professionals about significant risks. Amodei highlighted that within 6 to 12 months, AI-driven botnet swarms could emerge, potentially crippling the internet and causing hundreds of billions in damage, as reported by Tom's Hardware.
Meanwhile, companies like Nvidia and Palantir have proactively limited the use of advanced AI models due to concerns over data privacy and intellectual property rights, illustrating operational and regulatory challenges in AI deployment as detailed in another Tom's Hardware report.
This debate occurs amid intense scrutiny from legal experts and regulators about how best to govern AI. Legal professionals should monitor how industry preferences for self-regulation intersect with evolving compliance requirements and risk assessments within AI frameworks.
By the numbers:
- September 2, 2026 — Date of Nvidia CEO’s speech at G20 Innovation Ministerial
- 6 to 12 months — Timeline for potential AI botnet threat warned by Anthropic CEO
- Hundreds of billions — Estimated damage potential of AI-powered botnets as per Anthropic CEO
Yes, but: While Huang stresses engineering over lawmaking, significant AI risks like botnets underline the potential need for targeted regulation and stronger legal frameworks.
What's next: Global regulatory discussions on AI governance are expected to intensify ahead of the 2027 UN AI Summit, where balanced engineering and legal approaches will be key topics.