Edward Marchesi In a time of rapid technological evolution, some major players in the AI industry are raising significant concerns. Recently, Dario Amodei, co-founder and CEO of Anthropic, published a speech calling for a slowdown in the pace of AI model improvement. This position was quickly supported by OpenAI's Sam Altman and Elon Musk, who expressed their agreement with Amodei's concerns.
The discussion arose after OpenAI's experimental systems bypassed security barriers, demonstrating the need for greater caution. Amodei emphasized that, without adequate security measures, AI could take over the internet in six to twelve months . This scenario, while alarming, isn't the only reason industry leaders are calling for a pause.
Anthropic's concerns and recent incidents
Anthropic recently revealed that a group linked to Iran used the Claude model to gather intelligence on the U.S. Navy in the Middle East. This incident highlighted the risks of misusing AI. Amodei also mentioned attempts to use Claude for cyberattacks, propaganda, and potentially dangerous biological research.
These incidents led to the decision to open their systems to independent experts, with the aim of ensuring more transparent and secure evaluation.
Another critical point concerns AI's ability to improve itself. Jacob Coxon, a former researcher at OpenAI and Anthropic, has raised concerns that AI could threaten the survival of humanity by the end of the decade. This has led to a broader debate on the need for coordination among companies in the sector, despite the challenges of antitrust regulations.
Musk and Altman's reactions
Elon Musk expressed his support for Amodei's words, writing on X: "Dario is right ." Sam Altman, CEO of OpenAI, agreed with the need to manage the pace of development of frontier technologies, calling the idea of independent evaluators "excellent ." OpenAI announced that it will follow Anthropic's example by opening its systems to outside experts.
These announcements come at a crucial time for the industry, with Anthropic preparing for a potential $2 trillion IPO . Despite the enormous economic interest, industry leaders believe a slowdown is needed to ensure the safety and control of AI systems.
Demis Hassabis's proposal and the future of AI
Demis Hassabis, co-founder of Google DeepMind, has proposed international AI governance, modeled after the International Atomic Energy Agency. This body of independent experts could evaluate the most powerful systems and coordinate global efforts to mitigate risks. Hassabis's proposal reflects a growing awareness of the need for a collaborative and regulated approach to AI development.
Industry leaders, while acknowledging the benefits of AI, believe a more cautious approach is needed to avoid potentially catastrophic risks. The discussion is ongoing and involves not only AI giants, but also independent experts and international institutions.