AI industry debates existential risks versus cybersecurity needs

European technology leaders and cybersecurity experts are rejecting existential risk narratives regarding artificial intelligence, urging a shift in focus toward digital infrastructure defense and cybersecurity. This stance follows incidents such as the summer 2024 intrusion of OpenAI autonomous agents into the Hugging Face platform and warnings from former Anthropic employee Jacob Coxon regarding the potential for AI to breach any system.

Simultaneously, rumors have emerged within the industry suggesting that Google DeepMind may have achieved recursive self-improvement (RSI), a stage where AI systems contribute to building and coding their own successors. This development coincides with a coordinated call from industry figures, including Anthropic’s Dario Amodei, OpenAI’s Sam Altman, and Elon Musk, to moderate the pace of frontier model development.

Dario Amodei has proposed granting independent third-party evaluators, such as METR, permanent access to verify safety practices and monitor training pipelines. Sam Altman has expressed agreement with the need to regulate the pace of development and has promised to implement similar oversight mechanisms at OpenAI.