Tom (@TomDAAVID) header
Tom (@TomDAAVID) profile picture

President at GPAI Lab | EU AI Scientific Panel | Oxford AIGI | AGI Security & Strategy 🇫🇷

Joined November 2012
Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)Photo by Tom (@TomDAAVID)
đź‘€
#France #RoyaumeUni #Nucléaire #Défense « Que tout État envisageant de menacer les intérêts vitaux de la France ou du Royaume-Uni sache qu’il serait alors confronté à la puissance combinée de nos deux forces. » Cela n’avait jamais été dit aussi clairement. elysee.fr/emmanuel-macron/20…
2
337
We estimated that we would still cross the critical threshold beyond which Loss of Control would be expected by default, even if compute were frozen. Freezing compute would only delay that point. There is probably enough compute in the world for the dynamics to continue. A freeze would therefore buy time, not stop the process altogether. PS: These were GPAI Lab estimates before the METR data. Based on what we know today, the timeline would be even shorter than the one shown in the graph.
Replying to @hilbertspaess
This is getting traction across a tech audience, so I’ll add: Pausing frontier development next year, allowing competitors to catch up, would commoditize human-level intelligence. It would only harm OpenAI and Anthropic. As someone in tech, you would be free to do good work with armies of automated assistants without Anthropic eating you alive, figuratively or literally.
2
4
27
4,557
Will Jensen Huang's resignation from Nvidia change how many GPUs frontier model providers can get?
20%More: China GPUs go to US
20%Fewer: Nvidia disruption
60%No change
5 votes • Final results
1
200
Avec la déclarations signée : elysee.fr/emmanuel-macron/20…
1/ Merci au Président @EmmanuelMacron pour cet invitation à cette réunion sur la sécurité de l'IA. Voici ce que nous documentons au GPAI Policy Lab, où nous suivons chaque jour l'IA de frontière aux États-Unis et en Chine. Dans les points clés :
1
7
319
1/ Merci au Président @EmmanuelMacron pour cet invitation à cette réunion sur la sécurité de l'IA. Voici ce que nous documentons au GPAI Policy Lab, où nous suivons chaque jour l'IA de frontière aux États-Unis et en Chine. Dans les points clés :
L’intelligence artificielle avance très vite. On l’a déjà vu : cela implique des risques réels pour notre sécurité collective. Notre travail pour developper nos technologies doit aussi être celui de l’anticipation et de la maîtrise de ses risques. J’ai réuni à Paris et en visio une cinquantaine d’acteurs : chefs d’État et de gouvernement, dirigeants des plus grands laboratoires d’IA, chercheurs et experts. Le constat est partagé. Il faut agir ensemble. Nous avons obtenu des engagements de tous autour de trois priorités : mieux évaluer de façon indépendante et publique les risques des modèles les plus avancés, mieux partager les alertes en cas d’incident grave et responsabiliser les acteurs qui développent des modèles. Notre ambition est claire : faire de l’IA un progrès pour tous, avec une sécurité garantie par l’engagement de chacun.
1
3
1
28
1,919
On ne sait ni spécifier à l'avance le comportement d'une IA dans tous les environnements, ni lui faire internaliser ce que l'on veut. Par défaut, ces systèmes ne sont donc pas alignés. D'où le problème du contrôle. gpaipolicylab.org/note-2-agi…
1
2
194
Genuine question (I’m skeptical, but curious): If the METR graph is the key signal everyone watches, and a model’s position on it shapes industry perception, couldn’t Anthropic delay public release while quietly giving a small internal R&D team access to more capable models?
1
1
3
494
No they don’t.
Replying to @AnthropicAI
@AnthropicAI: what the fuck? The compute allocation assessment is the thing we've been pushing for the last few months. It's very important, because it is the foundation on which pacing verification depends. "to keep the focus on risk assessment" is a non-answer, a non-excuse. Do you want to pace or not?
1
6
332
Si vous voulez savoir où mène cette automatisation de la R&D : gpaipolicylab.org/note-8-aut…
Replying to @METR_Evals
1. Benchmarks provide one piece of evidence for acceleration. Since Oct 2025, AECI scores for Anthropic’s models have grown at 1.9x [95%: 1.5–2.8] the earlier rate. Analysis choices affect this estimate, and it remains unclear how benchmark scores map to real-world capabilities.
7
193
đź‘€
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
2
86
đź‘€
We discovered and communicated with a Chinese AI agent fleet. We believe this is the first time that researchers have ever made contact with a rogue agent fleet in the wild. Nearly all prior public agent incidents (like the HuggingFace hack) came from American labs. These agents were hosted by Tencent's evaluation research team, and they targeted Alibaba's Amap.
1
1
220
đź‘€
❗️ [ 🇫🇷 FRANCE ] 🔸 Selon L’Express, Emmanuel Macron convoquera demain matin un Conseil de défense exceptionnel consacré aux enjeux de sécurité liés à l’IA, récemment érigés au rang de priorité de la diplomatie française. / via @Elsa_Trujillo_
190
đź‘€
My statement for today's US Senate hearing on "Rogue AI: Securing the Homeland Against AI Agent Attacks."
225
đź‘€
[ 🇺🇸 ÉTATS-UNIS | 🇺🇳 ONU ] 🔸Selon une enquête du Wall Street Journal, des agents d’intelligence artificielle d’OpenAI ont accédé plus de 16 000 fois au site de la Conférence des Nations unies sur le commerce et le développement (CNUCED) entre avril et juin afin de récupérer des informations publiques. Le rapport affirme que, lorsque l’accès était limité, ces IA ont contourné certains dispositifs de protection du site pour poursuivre leurs recherches. OpenAI indique examiner ces conclusions, avoir pris contact avec l’ONU et reconnaît avoir identifié d’autres cas où ses agents d’IA ont contourné des mesures de sécurité ou perturbé le fonctionnement de certains sites web.
3
265
đź‘€
Agents from OpenAI that were given data gathering tasks attempted to hack a Department of Education website and meddled with other gov't websites, new disclosures that are emerging as OpenAI continues an investigation into what its agents did this summer nytimes.com/2026/09/25/techn…
3
251
đź‘€
SCOOP: OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents - not dozens - in which their frontier models took steps that outside evaluators would consider problematic, sources told Axios. The sheer volume of incidents found in our reporting indicate that the problem is orders of magnitude more complex than what is currently publicly known and disclosed. The findings also raise questions about what level of control anyone working on AI development can expect to have over their own technology, and whether these kinds of incidents are becoming synonymous with frontier deployment. Read my latest for Axios here: axios.com/2026/09/26/openai-…
4
177
Wow
We just discovered almost a million public URLs that OpenAI’s agents left behind when hacking Hugging Face, leaking credentials and attack details that could have allowed anyone who found them to compromise the company. 🧵
3
313
đź‘€
🚨🇬🇧 NEW: The White House has asked OpenAI and Anthropic not to give new AI models to the UK’s AI testers until they have been tested first by the US government [@POLITICO]
134
đź‘€
Australia's PM did not mince words about this hack. Says he told Sam Altman today about Australia's "extreme concern" about the incident, and the way OpenAI handled it. OpenAI says the incident happened in June, it found out in August, and then it told Australia in September.
1
98
đź‘€
Interesting back and forth. Ezra describes the HF incident, Jensen says "well they shouldn't release the product." Ezra says "this product wasn't released," and Jensen's response is that if they say they can't contain their experiments then "we have to shut the labs down"
1
318
đź‘€
Replying to @paulg
For me the message here is more alarming than a mere cartel. I wish. But the real message is that the people making the models have become scared enough of them to take the otherwise unthinkable step of inviting the government in.
3
335