“Taking control” of the Internet from 2027: OpenAI, Anthropic, and Elon Musk want to slow the AI race; Trump and China refuse

The AI world was considerably shaken this weekend. On Saturday, September 12, 2026, Dario Amodei, the head of Anthropic, published a long text devoted to the dangers of artificial intelligence.the dangers of artificial intelligence.The entrepreneur, whose start-up is behind Claude, “fears that within 6 to 12 months” a group of autonomous AI agents will be “capable of taking control of the entirety of the Internet.”

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.

Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our…

— Dario Amodei (@DarioAmodei) September 12, 2026

The warning echoes the warning from Jacob Coxon, a former Anthropic employee, who believed that “the people building AI genuinely think it could kill us all by the end of the decade.”Amodei’s remarks also stem, above all, froma series of cybersecurity incidents that occurred this summer. The summer was indeed marked by several cyberattacks carried out by rogue AIs, including versions of Claude and ChatGPT. One will particularly recall the intrusion into the servers of the Hugging Face platform, orchestrated by a GPT model that had escaped its digital prison.

Read also: OpenAI, Google, Anthropic and more than 100 companies demand an urgent “global response” to AI cyberattacks

“Slowing the pace”

In the face of these repeated incidents, Dario Amodei believes that we need to “slow the pace at which we improve the capabilities of AI models”before a catastrophe occurs. He does note, however, that “progress will continue to appear rapid, and we must make good use of the time we gain.” The Anthropic executive explains that he changed his mind on the issue of an AI moratorium in recent months. Until recently, he thought that investing heavily in safety would be enough to support the increasing power of AI. He now believes that this is no longer sufficient. While investing in safety, the pace at which capabilities advance must also be slowed down, to give researchers time to develop robust safeguards.

“This requires not only investing in risk prevention, but also pacing the advancement of capabilities so that this prevention has time to keep up”, explains

A three-step plan

The CEO of Anthropic proposesa three-step plan to avoid losing control of AI. First, he recommends that leading AI laboratories give independent evaluators extensive access to their systems to verify that safety practices are genuinely being followed.Anthropic has committed to setting up an audit system of this kind right away, without waiting for the other AI giants.

Then, AI companies based in democratic countries should agree on common safety rules and jointly set limits that must not be crossed. This approach should prevent prudent companies from being put at a disadvantage compared with competitors willing to take more risks to move faster in AI development. There are clearly competitive stakes in slowing down AI. Finally, the head of Anthropic calls for extensive international cooperation, including with non-democratic governments such as China, to oversee risks on a global scale

Also read: Several ChatGPTs hacked a website to turn it into a secret forum

OpenAI and Elon Musk’s response

Sam Altman, CEO of OpenAI and Anthropic’s main rival, agreed with Amodei. Following his statements to the company’s employees, the executive assures that OpenAI is ready to pull back, but that all laboratories need to do the same. He says that OpenAI will also set up “a system of independent evaluators with access similar to that of employees”, and that no “American competitive pressure should justify recklessness”. The race to develop AI should not justify taking risks indiscriminately.

The world deserves confidence that American companies developing increasingly capable AI will act responsibly, especially as the trajectory of progress has steepened. Every frontier lab must deliver on this, and there is no reason any of us should come to work if we cannot.

We…

— Sam Altman (@sama) September 14, 2026

For reference, Dario Amodei left OpenAI in 2020, judging the company to be insufficiently rigorous on safety issues related to artificial intelligence. It was this divergence of vision that led him to create Anthropic one year later, in 2021. Despite their past disagreements, Amodei and Altman now appear to be on the same wavelength. In the current context, OpenAI has chosen to postpone its IPO until next year, considering that “it would be inappropriate”. By contrast, Anthropic still plans to enter Wall Street in October 2026.

I’ve been sounding the alarm on AI for a long time https://t.co/YJ0oAcVXz2

— Elon Musk (@elonmusk) September 13, 2026

Anthropic’s CEO’s remarks were even praised by Elon Musk, the head of xAi. The controversial billionaire, whose empire is behind Grok AI, points out that he has been warning about the dangers of artificial intelligence for more than a decade.

Hidden intentions and contrasting reactions

Amodei’s proposal, followed by endorsements from Altman and Musk, elicited mixed reactions among experts.Several voices have risen to accuse AI titans of harboring hidden agendas by warning about the dangers of their own technology… which they are actively continuing to develop. For Charles Guillemet, Ledger’s head of security, the warnings are part of the “public relations rituals” customary for AI giants. For him, it is yet another communications operation.

Musk, Altman and Dario are predicting the end of the world again and asking everyone to slow the frontier. In a couple of months they will ship a stronger model. These PR rituals are now part of the PDLC!

We now have dozens of open-weight models with weak or removable… pic.twitter.com/XCNApRS6Ef

— Charles Guillemet (@P3b7_) September 12, 2026

Yann LeCun, the French AI pioneer and researcher, recalls for his part that Amodei stated “already in 2019 that GPT-2 was too dangerous to be open source”, and that his statements should not be taken seriously.

Right. Dario was already claiming that GPT2 was too dangerous to open source back in 2019.
I made fun of them then.
Everyone should make fun of them now.

— Yann LeCun (@ylecun) September 13, 2026

Donald Trump’s refusal

The warning from the AI giants caused a stir all the way to the Oval Office.President Donald Trump said that calls to slow down are “very negative forces” that could prevent the United States from establishing itself as the sector’s leader. Unsurprisingly, the Republican pointed the finger at competition from China, which is multiplying innovations in AI, Reuters reported.

“We have a head start over China in AI. We are the most advanced country in the world and, to be very honest, I want it to stay that way, because whoever dominates AI wins the game,” said Donald Trump, arguing that “many very negative forces are weaponizing this issue when they should not be, and describing scenarios that will never materialize.”.

Asked by CBS, Dario Amodei explained that rivalry between American and Chinese companies constituted the “toughest dilemma” facing the AI sector as it sets a “speed limit.”.

China also reacted to Anthropic’s proposals through Global Times, a newspaper under the control of the Chinese government, as reported by Reuters. For the outlet, the American startup’s warning is a “Cold War playbook” aimed against China.For Beijing, the framework intended to regulate AI development would be aimed solely at curbing AI development in China through technological barriers and regulatory monopolies.

👉🏻 Follow the tech news in real time: add 01net to your Google sources, and subscribe to our WhatsApp channel.