Startup CEO Backs Call to Slow AI Development

While most major artificial intelligence companies have remained silent following recent warnings about existential risks, Anton Osika, the chief executive of Lovable, has publicly broken that silence. In a recent post on LinkedIn, Osika expressed support for slowing down the development of potentially dangerous AI systems, stating that progress should pause when safety cannot be clearly demonstrated. He emphasized that decisions with such broad societal implications require public accountability rather than internal-only oversight.

The controversy stems from a warning issued by Jacob Coxon, a former researcher at OpenAI and Anthropic. Coxon accused the two leading labs of racing toward self-improving superintelligence without adequate safety measures. He described the situation as a serious gamble with human lives, asserting that the developers themselves believe the technology could pose a fatal threat within the next decade. This accusation has sparked a heated debate among researchers and policymakers about the pace of innovation versus the need for rigorous safety testing.

Founders push for international safety rules

Osika, who leads a Swedish startup focused on coding assistance, framed his support as a call for European and global leadership. He argued that the entities building the most powerful systems must be required to prove they can keep them under control. The CEO of Lovable, as reported by GN technics/ai (en-US), suggested that international coordination is necessary to prevent a scenario where competitive pressures override safety protocols. His stance highlights a growing demand among some tech leaders for transparent, externally verifiable safety standards.

This position aligns with earlier statements from other industry executives. Sam Altman of OpenAI has previously noted the need for collaboration with governments to resolve serious safety problems. Similarly, Dario Amodei of Anthropic warned about potential global security threats posed by advanced AI. However, Osika’s current statement is notable for its explicit endorsement of slowing development if safety is not met, a stance that goes beyond general calls for caution.

Researchers debate the likelihood of extinction

The intensity of the debate is reflected in the varying perspectives within the AI community. Evan Hubinger, who leads the alignment team at Anthropic, has personally expressed a belief that there is a significant chance AI could lead to human extinction within ten years. He admitted that the company does not yet have a complete solution for ensuring superintelligent AI follows human values. This admission underscores the uncertainty surrounding the long-term safety of these systems.

Conversely, other prominent figures in the field argue that such fears are exaggerated. Gary Marcus, an AI researcher, stated there is no realistic scenario for near-term human extinction. Melanie Mitchell, a professor at the Santa Fe Institute, similarly dismissed extinction fears as far-fetched. This split opinion illustrates the difficulty in reaching a consensus on the actual level of risk posed by current AI technologies.

Silence from major AI laboratories

Despite the public outcry, the leaders of OpenAI and Anthropic have not directly responded to Coxon’s specific accusations in recent days. Both companies have issued general warnings about AI risks in the past, emphasizing the need for careful management. However, their current silence has left a vacuum in the public conversation, allowing voices like Osika’s to gain prominence. The lack of direct engagement from the major labs has fueled speculation about their internal strategies and risk assessments.

The core of Osika’s argument rests on the idea that ambition for AI should be measured by what it makes possible for humanity, not just by speed or capability. He advocates for a model where safety is a prerequisite for advancement, rather than an afterthought. This trade-off between rapid innovation and guaranteed safety remains the central challenge for the industry as it moves toward more autonomous and powerful systems.