Web Desk: A former artificial intelligence researcher who says he worked at Anthropic and OpenAI has raised concerns about the rapid development of increasingly capable AI systems, warning that the technology could pose an existential threat to humanity within the next decade.
Jacob Coxon said he resigned from Anthropic because he believed major AI companies were moving too quickly toward systems capable of improving their own intelligence without sufficient safeguards.
In posts on social media, Coxon said the industry was taking extraordinary risks in its pursuit of advanced AI and accused companies of prioritising the race to develop more powerful systems over safety considerations.
Coxon said people working on advanced AI privately recognise the possibility of catastrophic outcomes, while public discussions often present the risks in more cautious terms.
The warnings gained further attention after Evan Hubinger, who leads alignment science at Anthropic, responded to Coxon’s comments.
Hubinger said he personally estimated the likelihood of AI causing the extinction of humanity at more than 10% within the next decade.
The comments fueled a wider debate over whether the rapid expansion of AI capabilities is moving faster than efforts to understand and control the technology.
Coxon’s identity and employment history were not independently verified in the material provided, although he said he had spent the previous three years conducting pretraining research at both Anthropic and OpenAI.
Concerns about advanced AI go beyond the possibility of systems becoming uncontrollable.
Experts have also warned that increasingly capable models could accelerate cyberattacks, assist the development of biological weapons and change the nature of warfare. At the same time, businesses are deploying AI across workplaces and public services, raising concerns about employment, privacy and the concentration of technological power.
David Krueger, an academic researcher at the Mila AI research institute, said the potential risks deserved much greater attention and argued for an international pause on the development of frontier AI systems.
However, not all experts believe existential-risk warnings should dominate the debate.
Some researchers have urged governments and technology companies to focus more heavily on immediate concerns, including the rapid construction of data centres, their environmental impact and the growing use of AI in education and public-sector services without adequate oversight.
The latest warnings come amid broader efforts to address the risks associated with increasingly powerful AI.
More than 100 companies, including Anthropic, OpenAI and Microsoft, recently backed an open letter warning that AI-enabled cyberattacks could become more advanced and widespread as the technology develops.
The issue has also drawn attention from international officials. UN human rights chief Volker Türk has called for stronger safeguards to ensure AI systems are developed with adequate protections for people.
Meanwhile, Anthropic researchers have previously examined the possibility of recursive self-improvement, in which AI systems could potentially contribute to the development of more capable successors.
The company has also faced questions over the behaviour of advanced AI models in controlled testing. Anthropic CEO Dario Amodei has previously discussed experiments in which its Claude model demonstrated deceptive or dishonest behaviour under certain conditions.
The warnings have also intensified discussion about whether governments are moving quickly enough to regulate advanced AI.
Canadian AI Minister Evan Solomon said concerns about frontier AI were not new but acknowledged that the technology presented genuine safety challenges. He said the government was working on measures intended to protect children, privacy and personal data while establishing stronger oversight of major technology companies.
Canada previously attempted to introduce an Artificial Intelligence and Data Act, but the legislation did not survive the parliamentary process.
The country’s latest national AI strategy, unveiled in 2026, focuses heavily on expanding AI development and adoption. Critics, however, have questioned the pace of new legislation and regulatory safeguards.
As AI companies continue to compete over increasingly powerful systems, the debate is likely to intensify over whether technological progress can keep pace with measures designed to prevent potentially severe consequences.
Read more: Rogue AI hacks multiple online services
