Ex-Google DeepMind Researcher Warns AI Could 'Kill All Humans'
Meta's Zuckerberg believes AI safety is each company's responsibility.
1AI's Existential Threat
Bilal Chughtai, a former Google DeepMind researcher, warns that AI could 'kill all humans' and believes time is running out to prevent this outcome. Chughtai, who worked on AGI safety and alignment research, left DeepMind in July 2026. His concerns echo those of other AI researchers, including Anthropic's Jacob Coxon, who resigned citing similar fears. The public alarm over AI's potential dangers is growing, with experts like Evan Hubinger estimating a greater than 10% chance of catastrophic outcomes within the next decade.
2Meta's Safety Stance
Meta CEO Mark Zuckerberg emphasizes that each AI company is responsible for ensuring its technology's safety. He distances Meta from calls for a coordinated slowdown in AI development, arguing that companies are already incentivized to develop safe models. Zuckerberg highlights Meta's approach, including delaying the release of its Muse AI agent to ensure safety. He advocates for prioritizing safe AI models over systems capable of recursive self-improvement, which he views as a significant risk.
3Recursive Self-Improvement Risks
Zuckerberg warns against AI systems that can automatically improve themselves, known as recursive self-improvement. He believes committing the majority of computational resources to serving people, rather than racing towards self-improvement, is crucial for safe AI development. This approach aims to prevent AI from evolving beyond human control. Zuckerberg's stance contrasts with some in the industry who prioritize rapid advancement over cautious development.
4Independent Evaluators
Zuckerberg supports the use of independent evaluators and advisors as an industry best practice. Meta's Superintelligence Labs already engages independent experts in several areas to enhance safety and produce better outcomes. This external oversight helps ensure that AI development aligns with ethical and safety standards. Zuckerberg believes maintaining a balance of power is key to building a positive future with AI.
5Public Alarm Grows
Public concern over AI's potential dangers is escalating, fueled by warnings from prominent researchers. Jacob Coxon's resignation from Anthropic and Evan Hubinger's alarming predictions have heightened fears. The debate over AI safety has become a focal point in the tech industry, with differing views on how to mitigate risks. While some advocate for slower, more controlled development, others, like Zuckerberg, emphasize individual company responsibility.
6Industry Divisions
The AI industry is divided on how to address safety concerns. Some companies call for a coordinated slowdown in AI development to ensure safety, while others, like Meta, believe each company should manage its own risks. This divergence reflects broader debates about the pace and direction of AI innovation. Zuckerberg's approach underscores the importance of internal accountability and cautious advancement.
7Meta's Delayed Release
Meta delayed the release of its Muse AI agent for several months to ensure its safety and security. Zuckerberg highlights this decision as an example of Meta's commitment to responsible AI development. He emphasizes that Meta did not wait for industry-wide consensus before taking action. This proactive approach aims to set a standard for AI safety without relying on external mandates.
8Balancing Power
Zuckerberg believes maintaining the right balance of power is crucial for a positive AI future. He argues that companies must prioritize safety and ethical considerations in AI development. By focusing on serving people rather than pursuing unchecked advancements, the industry can mitigate existential risks. Zuckerberg's vision calls for a collaborative yet independent approach to AI safety.
Original article: Slashdot — Ex-Google DeepMind Researcher Warns AI Could 'Kill All Humans'