Concerns over artificial intelligence safety are growing as the technology advances and questions emerge over whether safeguards can keep pace. Microsoft co-founder Bill Gates, Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and SpaceX CEO Elon Musk have all warned about the potential risks posed by increasingly powerful AI systems.
The debate has intensified in recent weeks. An AI researcher who left Anthropic accused both Anthropic and OpenAI of “gambling with our lives.” Later that day, an Anthropic alignment lead warned that there was a greater than 10% chance AI could “kill all humans” within the next decade.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
Trump Emphasises Competition Over AI Risks
President Donald Trump has taken a different approach, repeatedly downplaying the risk of AI-driven extinction and framing the technology primarily as a strategic competition with China.
His administration recently established a new “AI Force,” led by Director of National Intelligence Jay Clayton, to support the AI industry and identify bad actors. The initiative reflects the administration’s focus on strengthening the United States’ position in the rapidly developing sector.
Nadella Calls For AI Systems To Be Observable
Microsoft CEO Satya Nadella has argued that the next generation of AI systems should be built around what he calls “principles of observability,” allowing their behaviour to be monitored, tested and independently assessed.
His proposed safeguards include model diversity, human-readable records of model actions, continuous system testing, independent controls, auditability, containment measures and incident disclosure. Together, these mechanisms could help organisations identify problems, understand how AI systems behave and respond when something goes wrong.
Nadella’s argument centres on the idea that safety should not depend solely on confidence in a particular model. Instead, systems need safeguards that make their behaviour transparent and allow people to retain meaningful oversight.
“The most trustworthy Super Intelligence system will not be the one with the model we trust most. It will be the one that enables us to trust the model the least,” Nadella said.
