
Safety @OpenAI Prev: PM of Gemma @GoogleDeepmind \ PM in Responsible AI @GoogleAI Side: adjunct. prof @sciencespo & @ESCPeurope
Joined May 2011
- Tweets1.1K
- Following1.3K
- Followers1K
- Likes4.1K
Extremely interesting finding, thanks for sharing @cohere
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
65
This is so interesting : to the many thinking Europe lags behind because of regulations.. this chart offers another read: E. lack capitals (or lacks the financial ways to channel its capital to prodctive and innovation-driving projects)
Ludovic Peran retweeted
Underrated fact that three out of three of the most cited AI researchers of all time have all been very vocal about existential risk from AI.
Are they part of the IPO marketing scheme too?
Ludovic Peran retweeted
Julie is amazing and her position on slowing down is consistent with our chief scientist’s essay. I also think some pacing is likely to be necessary. (Speaking personally and not in my capacity as a janitor.)
Ludovic Peran retweeted
The first autonomous agent cyberattack is an unprecedented event that deserves unprecedented transparency. Today we’re sharing everything we can: a full technical timeline, an interactive replay, and how we used an open model to defend ourselves, so defenders everywhere can learn from it and prepare for what’s next.
huggingface.co/blog/agent-in…
Ludovic Peran retweeted
We're excited to announce that Resolution has a $160M grant from Coefficient Giving: $108M unconditional, with a further $52M conditional on hiring and compute needs. We'll use it to grow teams across our research portfolio and invest heavily in research automation. 🧵
Interesting move: Beijing is looking at curbing overseas access to China's top AI models, sources say
reuters.com/world/beijing-is…
1
83
Ludovic Peran retweeted
I’m excited to share that I’ll be joining OpenAI and look forward to working with the exceptional team there.
It was a difficult decision to move on. I’m incredibly proud of the amazing team at Google and everything we’ve built together. It has been an honor and a pleasure to work with all of you.
Introducing the OpenAI Safety Fellowship, a new program supporting independent research on AI safety and alignment—and the next generation of talent.
openai.com/index/introducing…
Introducing GPT-5.5
A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.
Now available in ChatGPT and Codex.
Ludovic Peran retweeted
We’re releasing prompt-based teen safety policies for gpt-oss-safeguard.
They’re designed to help you identify and moderate teen-specific content, and turn safety requirements into classifiers for real-time filtering or offline analysis.
openai.com/index/teen-safety…
that's a big milestone: looking forward to know what our European community is going to build with this 🩶 - share you best European use cases!
Three million people are now using Codex weekly - up from two million a little under a month ago. Incredible to see the growth. Thank you to all of you and to the ecosystem we’re part of. To celebrate, we’re resetting rate limits so you can keep building, and we’ll reset them every additional 1M users until we reach 10M, so we can keep celebrating along the way.
Enjoy and thank you!
1
128
Building for teens? Check out our Teen Safety Policy Pack and contribute to it --> github.com/openai/teen-safet…
We’re releasing prompt-based teen safety policies for gpt-oss-safeguard.
They’re designed to help you identify and moderate teen-specific content, and turn safety requirements into classifiers for real-time filtering or offline analysis.
openai.com/index/teen-safety…
1
112
Open question to the safety community: are there teen-specific safety classifiers for content detection?
Roblox Guard has a 'Child Exploitation' category.. Anything else?
1
111
Join us in SF on Dec 8 for a special hackathon, ROOST and @huggingface
🧑💻 gpt-oss-safeguard Hackathon 🧑💻
Join us Dec. 8 in SF for the Open Safeguard Hackathon — a collaborative event by OpenAI, ROOST & @HuggingFace to explore how open models can shape safer digital spaces and explore the future of open-weight reasoning and online safety.
Apply to participate: events.openai.com/gpt-oss-sa…
189
Looking forward to iterate on gpt-oss-safeguard further !
Provide feedback through ROOST and its new model community github.com/roostorg/open-mod…
Now in research preview: gpt-oss-safeguard
Two open-weight reasoning models built for safety classification.
openai.com/index/introducing…
2
269



























