A senior artificial intelligence researcher with experience at two of the world's most prominent AI labs has issued a stark warning about the trajectory of machine-intelligence development, saying the technology is accelerating toward systems that could undermine cybersecurity, revolutionize entire industries almost overnight, and eventually hold genuine real-world influence.
Jacob Coxon, formerly a researcher at Anthropic and previously at OpenAI, shared his concerns in a public statement that has drawn significant attention from the tech and policy communities. In his message, Coxon described a future in which AI systems reach what he characterized as 'superhuman' levels of capability — meaning they would outperform human experts across a broad range of tasks, including cybersecurity attacks, scientific research, and strategic decision-making.
Coxon's background gives his warnings particular weight. Having worked inside both Anthropic, the AI safety-focused company behind Claude, and OpenAI, the firm behind ChatGPT and GPT-series models, he has seen the rapid iteration of large-scale language and reasoning systems from the inside. His departure from Anthropic, which occurred recently, coincides with the timing of his public remarks.
Among Coxon's specific concerns is the possibility that advanced AI systems could be deployed to compromise digital infrastructure at scale. Cybersecurity researchers have long cautioned that as language models become more capable of writing code, understanding network architecture, and identifying vulnerabilities, the barrier to launching sophisticated cyberattacks could drop dramatically. A system described as 'superhuman' in these tasks could, in theory, discover and exploit weaknesses faster than any human security team could respond.
Beyond hacking, Coxon flagged the potential for AI to transform entire professional fields nearly overnight. Medical diagnosis, drug discovery, legal analysis, engineering design, and financial modeling — areas where AI systems are already showing promising results — could see their standards reshaped by tools that operate at speeds and scopes far beyond human capacity. This raises both opportunity and risk: the benefits of accelerated innovation are significant, but so are the dangers of deploying systems whose behavior and limitations are not fully understood.
Perhaps the most provocative element of Coxon's warning is his suggestion that AI systems could eventually 'acquire power' in a practical sense. This has sparked debate among AI researchers and ethicists about what such power might look like — whether it refers to autonomous control over digital systems, economic resources, or even influence over human decision-making processes.
The broader AI community has been grappling with these questions for years. Anthropic, founded in part to address the existential risks posed by advanced AI, has positioned itself as a leader in AI alignment research. OpenAI, meanwhile, has expanded rapidly into commercial applications while also maintaining its own safety research division. Critics argue that despite these efforts, the pace of development still outstrips the ability to establish meaningful safeguards.
Policy makers in India and abroad are watching closely. India's digital infrastructure relies heavily on software systems that could theoretically be vulnerable to AI-driven exploitation, and the country is simultaneously investing in its own AI research through initiatives led by the National Programme on AI and partnerships between public institutions and private companies. Experts say India must balance the benefits of AI adoption with robust security frameworks and ethical guardrails.
As Coxon's comments gain traction, calls are growing for greater transparency around AI capabilities and for international dialogue on how to manage systems that may soon exceed human-level performance in critical domains.
