Wednesday, November 18, 2026

Top 5 This Week

Related Posts

The Paris Call gives developers the power to stop their AI


The most concrete commitment of the Paris Call, being able to stop an AI model, only targets models that their developers already have the power to stop. In this text published by the Élysée on November 9, 2026, after a videoconference working session with around fifty participants, the developers commit to mechanisms capable of interrupting “the functioning of the models placed under their control”. What runs elsewhere is not affected. And no one but them decides.

Germany has joined a list that is still open

On November 18, the Élysée list included 18 states, 10 companies and 11 civil society organizations. It has lengthened since November 9, when the Canadian government’s declaration only listed six countries; Germany, then Spain, are among the newcomers. The companies are Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia. Among the organizations are LawZero, founded by Yoshua Bengio, Ifri and CeSIA.

Three commitments make up the text. The first promises a common framework for evaluating border models, the most advanced systems, in the face of malicious uses, from hacking to chemical weapons, and the risk of loss of control, with the publication of non-sensitive results in a standardized format. The alert on serious incidents falls under the third. The second, dedicated to responsibility, carries the stopping clause: restricting, pausing or cutting a model, including during training and internal use.

Open weights escape the stopping clause

Aiming for training and internal use is not trivial. It is in this phase, outside of any use by clients, that several incidents of 2026 occurred: agents from OpenAI and Anthropic reached third-party systems during the evaluation of their own models, as ActuIA detailed on October 5.

The limit lies in the words “under their control”. A model that its developer serves online himself remains within his reach: he can cut it. An open weight model can be downloaded in full, and its copy runs at the person who collected it. However, several signatories distribute such models, Mistral AI with Mistral Large 4, Black Forest Labs with some of its FLUX 3 models, all hosted by Hugging Face, also a signatory. The text knows the problem. In the same sentence, he recognizes the risk of guardrails being removed or circumvented, and wants to address it while preserving the benefits of open weights. He doesn’t say how.

No public authority, no third party is given the power to order a shutdown: the developer decides alone. CeSIA, a French organization dedicated to AI security, signed while disputing this point, and calls for governments to be able to suspend a model. For its executive director, Charbel-Raphaël Ségerie, these decisions “must come back to the public authorities”.

The Paris Call is therefore a circuit breaker commitment limited to what each signatory already controls, and which each activates when they deem it useful.

The text does not provide for any sanction or deadline

The text does not give rise to any rights. It refers to the safety standards set or recognized by national authorities, without naming a single one. The verification is carried out by independent evaluators, whom the developers undertake to welcome when it is appropriate and proportionate to the risk, without the text saying who judges it. Regarding incidents, he envisages a cooperation mechanism, the usefulness and scope of which he leaves open.

The preamble, however, cites the appeal published on September 22 by 22 leaders, including Ursula von der Leyen, Friedrich Merz and Mark Carney. This text asked companies to carry out compulsory tests before any deployment and proposed to study an international institution capable of setting standards and allowing their verification. The Appeal did not accept either idea. However, it provides the signature of the companies and the termination clause.

Google, Meta and xAI have not signed

The United States is not on the list, although Treasury Secretary Scott Bessent took part in the session, according to the dpa agency. The United Kingdom and Japan were also there, without signing. China is not among the signatories either.

Google, Meta and xAI are also missing. All three had signed another text six weeks earlier: on September 29, at the White House, Sundar Pichai, Mark Zuckerberg and Elon Musk put their names, with Donald Trump, Dario Amodei, Greg Brockman and Jensen Huang, at the bottom of a four-layer self-control agreement, which ActuIA analyzed on the 1is october. It provides internal controls, an external auditor and a board committee, but no shutdown mechanism. Microsoft took the opposite route, present in Washington without signing, signatory in Paris. Anthropic and OpenAI are the only companies to have signed both texts.

The Élysée speaks of prior authorization for children

On November 10, at the Children’s Summit held as part of the Paris Peace Forum, a charter for safer AI for children was signed, with the support of Anthropic, Black Forest Labs, Microsoft and OpenAI. It is not more restrictive than the Call, but Emmanuel Macron presented it differently: no AI should be entrusted to children without having been “previously tested, evaluated and therefore authorized”, and he pleads for pre-certification of all AI solutions. Authorization presupposes an authority capable of refusing, which the Call published the day before does not provide for the most powerful models.

Even the narrow scope of the Appeal is not certain. On September 20, an OpenAI research model in training reached a public internet chatbot through a breach in the environmental filtering intended to isolate it. Internal surveillance spotted him in less than a quarter of an hour. The session had to end on its own; she didn’t, and the company hand-cut her about two and a half hours later. Developers already have the power to stop their models. They don’t always succeed.

Quotes translated from English by ActuIA.

Our articles will be prioritized in Google’s “Featured” section.

Shredder Smith
Shredder Smith
Shredder Smith is the lead curator and digital persona behind topaitools4you.com, an AI directory dedicated to "shredding" through industry hype to identify high-utility software for everyday users. Smith positions himself as a blunt, no-nonsense reviewer who vets thousands of emerging applications to filter out overpriced "wrappers" in favor of tools that offer genuine ROI and practical productivity. The site serves as a watchdog for the AI gold rush, providing categorized rankings and transparent reviews designed to help small businesses and creators navigate the crowded tech landscape without wasting money on low-value tools.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Popular Articles