September 2026 will be remembered as the month of grand technological drama. Between warnings from researchers claiming that current models are approaching a point of uncontrollable autonomy and statements from U.S. lab directors predicting that AI could take control of the Internet “within a year,” Silicon Valley is playing its favorite tune: metaphysical fear.
After years of designing, deploying, and auditing artificial intelligence systems in real-world environments, our assessment at Deeplink is clear: this sensationalism is a business model, not a scientific warning. Here’s our take.
The marketing smokescreen: selling the end of the world to cover up abuses
As a compelling analysis in the newspaper Le Temps1 recently pointed out, the tech industry has invented a unique marketing paradox: selling a product by claiming it will destroy the world in order to drive up its stock price.
By raising the specter of an uncontrollable superintelligence, the San Francisco leadership is pulling two powerful psychological levers:
- Proof by threat: Claiming that an algorithm is on the verge of taking over the global network is a way to convince investors and boards of directors that it has immeasurable economic value.
- The Regulatory Smoke Screen: While policymakers and ethics committees debate Hollywood-worthy scenarios, public attention is diverted from the immediate abuses: the massive plundering of protected data without consent, the colossal environmental impact of the infrastructure, and the erosion of privacy.
Crying wolf after having disabled the safety locks yourself isn’t rocket science. It’s commercial cynicism and a strategy to shift collective responsibility.
The scientific diagnosis: risks associated with use, not agency
The primary danger of this sensationalism is that it amounts to pure science fiction. We are still a long way from Skynet. The International Report on AI Safety, published in February 2026 under the chairmanship of Yoshua Bengio—one of the world’s most alarmist researchers on these issues—is unequivocal: “Current AI systems do not pose any immediate risks of loss of control.” To date, a general-purpose AI is still unable to carry out an autonomous project lasting several days, remain reliable without hallucinating, or reason about unprecedented scientific problems.
Today’s AI is not a living entity; it is a tool. But a tool used improperly can cause major damage, just like a hammer: no one accuses a hammer of having hostile intentions, and yet, if used improperly, it can be fatal.
In our day-to-day work at Deeplink, we see that the real risks are already there, measured and documented:
- Loss of sovereignty: The growing dependence of European governments and companies on U.S. AI infrastructure that they neither own nor control.
- Automation bias: A proven human tendency to blindly follow machines. A clinical study2 involving nearly 2,800 participants demonstrated a 6% decline in doctors’ ability to detect tumors after several months of AI-assisted diagnosis.
- Emotional dependence: The widespread deployment of “AI companions” is creating documented signs of psychological dependence among tens of millions of users, to the point that China is already banning some of these apps.
None of these examples requires “superhuman” AI. These are risks related to usage and scale, not risks related to algorithmic autonomy. Treating deepfake fraud or the impact on employment as harbingers of a robot apocalypse is missing the point. It prevents us from taking the concrete and immediate measures (identity verification, labor market regulation, technical audits) needed to manage the real world.
Regulatory capture: Europe writes laws, America codes the future
This strategy of regulatory chaos directly benefits the tech giants. The Old Continent, historically fond of legislation, has a predictable reaction to fear: it reaches for its pen to draft complex laws and organize ethics committees, rather than for its keyboards to code the AI of tomorrow. By portraying AI as an unstoppable force, tech companies are pushing for regulations that stifle local competition and leave them with extraordinary free rein.
Choose Mastery Over Fear
The future of our digital economy must not be dictated by players who simultaneously peddle both the threat and the remedy. Businesses and institutions do not need prophets of doom. They need technology partners who understand exactly how the technology works, who take engineering responsibility for it, and who protect both their security and their sovereignty.
In the face of fear-mongering and Hollywood fantasies, we stand firm with the certainty of mastery.