Anton Osika, the CEO of the Swedish startup Lovable, stated on Thursday, Sept. 10, 2026, that he supports slowing the development of artificial intelligence in cases where safety cannot be proven. His comments followed a public warning from a former researcher at OpenAI and Anthropic regarding the risks of developing superintelligent systems.
The discussion was prompted by Jacob Coxon, a 27-year-old researcher who resigned from Anthropic during the week of Sept. 7. Coxon alleged that major AI laboratories are competing to create self-improving superintelligence without acting responsibly. On Wednesday, Sept. 9, Coxon wrote that many developers believe AI could pose an extinction-level threat by the end of the decade, a claim he characterized as a serious warning rather than a marketing tactic.
In a LinkedIn post, Osika argued that decisions impacting the general public require accountability and that developers of powerful systems should be required to demonstrate control over them. He expressed support for international coordination to manage dangerous development. While some researchers, such as Gary Marcus and Professor Melanie Mitchell, described near-term extinction scenarios as unrealistic or far-fetched, Evan Hubinger, who leads the alignment team at Anthropic, stated he personally believes there is a more than 10% chance AI could lead to human extinction within ten years.
The scale of the concern is measured by the timeline provided by researchers, who suggest these risks could materialize by the end of the decade. If the calls for slowing development are adopted through international coordination, citizens might notice a change in the pace at which new AI tools are released to the public or changes in how the technology is regulated.
The next steps involve how major AI labs and international regulators respond to these calls for safety demonstrations. While OpenAI CEO Sam Altman previously mentioned in April that the company might need to work with governments to resolve safety problems before proceeding, neither OpenAI nor Anthropic issued a new statement in response to the latest claims. Future actions could include legislative hearings or the development of new international safety standards for "vibe-coding" and superintelligence research.