Adel Ka
@0x4D31
security engineer @openai | ex-google. views are my own, not my employer's!
Microsoft Threat Intelligence uncovered a macOS vulnerability, tracked as CVE-2025-31199, that could allow attackers to steal private data of files normally protected by Transparency, Consent, and Control (TCC), such as caches used by Apple Intelligence. msft.it/6015sHUoS
normal day in detection & response: cognitive overload, nonstop context-switching. brains weren’t made for this, hah, but i enjoy pushing mine. 🤪
just another week at openai!
To summarize this week: - we released general purpose computer using agent - got beaten by a single human in atcoder heuristics competition - solved 5/6 new IMO problems with natural language proofs All of those are based on the same single reinforcement learning system
Gold medal-level performance on the 2025 International Math Olympiad from our latest experimental reasoning LLM. Model operated in natural language (i.e. outputs natural language proofs) under the same rules as humans (e.g. 4.5 hours per session, no tools). Amazing milestone!
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).
We competed live in the @atcoder World Finals Heuristic Contest and came in 2nd! Models like o3 rank among the top-100 in coding / math contests, but AFAIK this is the first top-3 placement in a premier coding / math contest. There's a huge gap between top-100 and top-3!
Congratulations @FakePsyho on a nail-biting performance! Great showings as well from @bminaiev, @andresnds, and @_lorenzkuhn representing OpenAI. It’s been fantastic sponsoring AtCoder World Finals @atcoder. We’re excited to share some of the model solutions with the world.
the new chatgpt agent is so cool! super satisfying to watch it use multiple tools to get things done. perfect release timing too.. i’ll use it to prep my slides for dc33, and for once, not be making last-min tweaks before my talk. openai.com/index/introduc…
ChatGPT can now do work for you using its own computer. Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths.
Agent Security is hiring! We are @openai's dedicated team focused on securing our agentic AI systems - we operate where traditional appsec/infrasec boundaries blur. This is frontier work: openai.com/careers/securi…
excited to be back in vegas for my second @defcon, and second talk on the creator stage! i’ll be diving into a mix of my favorite things: network fingerprinting, honeypots, and ai agents! tool drops in a few days--stay tuned 👾defcon.org/html/defcon-33… @wallofsheep
got a few more pet projects to open-source—more interesting than the ones i’ve shared. if only we had more company-wide week offs! 😅
lol! 'positive review only': researchers hide AI prompts in papers! asia.nikkei.com/Business/Techn…
So apparently the --host option in sudo has been broken since 2013. You could just trick it into accepting remote rules on the local system and get root. No exploit needed. And nobody noticed. For 12 years. Open source security at its finest: “If enough people look at the code,…