Amid a flurry of alarming statements from leading artificial intelligence companies, a prominent French AI researcher is urging the public to remain calm and skeptical of what he describes as strategic fear marketing designed to serve corporate interests.
A Pattern of Alarmism
In recent weeks, a wave of dire warnings about AI has dominated media and social networks. Among the most striking was the resignation of Jacob Coxon, a researcher who left OpenAI for Anthropic, accusing both companies of “playing with our lives.” His departure was followed by announcements from OpenAI, Anthropic, and xAI suggesting they were prepared to slow down AI development.
Laurent Daudet, a professor at the École Normale Supérieure (ENS) and founder of the generative AI startup LightOn, offered a counterpoint to this narrative in an interview with BFMTech. He cautioned against taking such statements at face value.
“We need to keep a cool head and step back from these remarks,” Daudet said. He pointed to a precedent: in 2019, when OpenAI released GPT-2, the company claimed the model was “too dangerous to make public.”
The Business of Fear
According to Daudet, fear marketing benefits the industry’s largest players. “It allows companies developing these models to become increasingly secretive about how they are designed and deployed, while suggesting they have an insurmountable lead over competitors,” he explained.
He predicted such declarations could multiply as Anthropic and OpenAI prepare for stock market listings, and as competition between China and the United States intensifies.
Real Risks Demand Real Attention
Daudet warned that apocalyptic discourse should not obscure concrete challenges. These include AI’s impact on employment and education, as well as the growing environmental cost of the technology, which relies on energy- and water-intensive data centers.
With AI development still largely unregulated, Daudet argued it should be subject to “more genuinely independent audits,” with methodology overseen by a broader range of stakeholders—including open-source developers, civil society, and academics—not just private companies.
He listed tangible risks: cybersecurity threats, disinformation, job displacement, climate impact, educational disruption, economic imbalances, and technological dependency.
Skepticism Toward Superintelligence
However, Daudet expressed doubt about the prospect of a rebellious superintelligent AI. “We are still very far from anything resembling ‘general AI,'” he said.
“Even though I love science fiction, we must be wary of the fantasies it induces, particularly anthropomorphic ones that attribute feelings, intentions, or consciousness to an algorithm,” he noted. “There is always a human who installed the model, a physical server running it, and another human who launched it on a set of instructions.”
Still, he acknowledged room for improvement in AI “alignment”—the process of defining a system’s intentions and values. This parameter, harder to program than a pure mathematical formula, can contain imperfections leading to serious errors.
