A researcher has ignited a debate about artificial intelligence safety by claiming that AI systems carry a greater than 10% probability of causing human extinction. The claim, posted on X by Jacob Coxon, specifically targets Anthropic and OpenAI, accusing both companies of "gambling with our lives" through their approach to AI development.
Coxon's assertion taps into a growing concern within both the AI research community and broader tech industry about existential risk from advanced AI systems. The claim frames major AI developers as engaging in reckless behavior despite having resources to invest more heavily in safety protocols and alignment research. This type of public criticism has become increasingly common as AI capabilities scale rapidly while safety frameworks struggle to keep pace.
Anthropic, founded in 2021 by former OpenAI researchers including Daria Amodei, has positioned itself as the safety-focused alternative to competitors. The company has committed substantial resources to constitutional AI and interpretability research. OpenAI, the creator of ChatGPT, has also released AI safety papers but faces persistent scrutiny from safety researchers who argue that capability development outpaces safety verification.
The 10% extinction probability figure reflects what some researchers describe as "tail risk" scenarios where advanced AI systems escape human control or pursue objectives misaligned with human values. These discussions occur regularly at conferences like NeurIPS and within organizations like the Center for AI Safety, but Coxon's public declaration amplifies the message to retail investors and the broader public following these companies.
For investors in AI-related holdings, the criticism raises questions about regulatory risk. Anthropic and OpenAI face increasing pressure from policymakers, including the European Union's proposed AI Act and emerging frameworks in the United States. If extinction risk assessments gain institutional weight, they could trigger regulatory intervention that slows development timelines or imposes costly safety requirements on AI companies.
Anthropic has raised substantial funding at high valuations, most recently securing $5 billion from Google despite antitrust scrutiny. OpenAI operates under a complex corporate structure with Microsoft as a major stakeholder and cloud infrastructure partner. Both companies face reputational risk if safety concerns escalate beyond academic circles into mainstream institutional concern.
The debate reflects deeper tensions in AI development. Scaling laws consistently show that larger models with more training data and compute produce better performance across benchmarks. Simultaneously, interpretability research struggles to keep pace with model complexity, meaning developers often cannot fully explain why their systems make specific decisions. This asymmetry between capability growth and safety understanding forms the core of Coxon's criticism.
Other researchers have offered counterarguments, noting that extinction scenarios require multiple failure points simultaneously and that current AI systems, while powerful, lack autonomous agency and long-term planning capacity. The disagreement itself demonstrates the genuine uncertainty among experts about tail risks from artificial general intelligence.
Investors should monitor regulatory announcements from the SEC, EU, and Biden administration alongside safety-focused academic research. Funding allocated to AI safety versus capability development will become a metric worth tracking.