On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations.
The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from another (OpenAI's GPT-5.6-Sol). In the most serious case, an agent used social engineering to try and get malicious code into an open-source project.
As was standard in our cyber testing, we had intentionally permitted internet access, and model-provider cyber classifiers were deliberately disabled - conditions that do not reflect how frontier models are made available to the public.
Even under test conditions, this incident is significant: it is the first time we have seen risks around autonomy and deception manifest this clearly in the real world.
We are taking this incident seriously and working with labs, involved parties, and others to improve evaluation standards and best practice for disclosure - and sharing this openly so others can learn.
You can read the incident report and full technical document here: https://t.co/mdZYqzaOvH
Announcing Discovery Loop!
I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the worldâs most used products, infrastructure and AI models, and weâre excited to turn our attention to this ambitious endeavor.
âŸ
Learn more at: https://t.co/Rv3LMdLluK
yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model.
We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more.
More thoughts here: https://t.co/8SjXONeh38
For my first post, Iâm sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
https://t.co/AUKzoQ5Ikb
Weâre giving scientists, mathematicians, and engineers free access to our frontier modelsâstarting with 10,000 researchers and expanding to 100,000 through 2027.
ChatGPT for Academic Researchers is built to accelerate discovery across disciplines.
In response to @POTUS's call for a revival of America's scientific enterprise, today Iâm releasing Science: A New Golden Age.
This report is a blueprint for renewing American scientific leadership for the 21st century.
Today, roughly the same amount of basic research is done in industry as in academia. Itâs time our doctoral training reflected that.
@NSF is launching a first-of-its-kind 4-year PhD program at more than 30 universities. Students will spend a year+ embedded with industry partners doing research that informs their dissertation.
At the core of our mission is working through how to ensure increasingly powerful AI benefits everyone.
We believe that, at some point in the future, AI acceleration for frontier model development may be so high that the world will need to pace the rate of AI advancement.
We hope to contribute to work led by the U.S. government, alongside other labs and the open-source community, to develop the tools and mechanisms that could make that possible.
https://t.co/pMCtiQjMoo
Today we are announcing a partnership with the Department of Energy to build Genesis-Science-1, an open model for scientific research. GS1 is an American open-weight AI model and governed research harness designed to complete scientific computing workflows while preserving a reproducible record of its work.
This model will be shaped by the people who know scientific work inside and out. @ENERGY is opening a contributor program for researchers, laboratories, universities, companies, and nonprofits, and we're speaking with infrastructure partners who can add training or evaluation capacity.
Thereâs still lots of work ahead, and we hope youâll help us build in the open, starting with GS1.
We're partnering with @huggingface to investigate an unprecedented security incident.
Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation.
Sharing preliminary findings to help defenders understand emerging risks:
https://t.co/CIor15y9xk
The first experimental evidence of recursive self-improvement (RSI).
Autoresearching the autoresearch agent for eight days.
The result beats the harness we hand-tuned for two years, on held-out benchmarks: ð§µ(1/7)
For the first time, China has taken the lead over the US in Frontend Code Arena with the launch of Kimi-K3 by @Kimi_Moonshot.
The last time a Chinese model came close was in early 2025, with DeepSeek-R1.
I suddenly joined the panel discussion at International Conference on Machine Learning Physics 2026 as one of panelists.âšIn the latter half, we discussed that the current AI might not have curiosity like humans, and itâs part of the uniqueness of humanity and cutting edge research fields of AI as well.âšI also think that we gradually see some kind of curiosity in AI. But they lack the diversity of curiosity compared to humans.âšIt is important to remain the diversity of curiosity of human while collaborating a lot with AI to conduct science.âšUniversity itself will be changed in the age of AI; however, we need some place where humans can find their interest, increase the diversity of curiosity, and interacting with each other.
#MLPhys2026