OpenAI 详解前沿模型训练环境的安全加固与红队测试
模型发布/更新
-
Moonshot AI releases Kimi K3.1 Pro with 1M token context, open weights due by November 4
Moonshot AI released Kimi K3.1 Pro, the first model in the Kimi K3.1 series, a 3T-class open-weight model with a 1…
-
MiniMax releases M3.1 Flash with nearly 3x faster reasoning and 70% lower usage cost
MiniMax released MiniMax M3.1 Flash. The anonymous model Space Bunny was an early version of it and had processed more…
产品发布/更新
-
DeepSeek Fully Open-Sources Internal Inference Framework DeepInfer
DeepSeek announced that its internal inference framework DeepInfer is now fully open-source, alongside the…
-
Anthropic opens Claude Duet research preview, with Claude taking over daily work as a proactive agent
Anthropic opened the Claude Duet research preview, a standalone app that lets Claude work as a proactive agent, now…
-
Google and Meta release TorchTPU: a native PyTorch backend for TPU
Google and Meta jointly open-sourced TorchTPU, a native PyTorch backend for Google TPU that lets models run out of the…
行业动态
-
OpenAI details security hardening and red teaming of frontier model training environments
OpenAI describes security hardening measures for its frontier model training and evaluation environments, including no…
-
OpenAI discloses Patch the Planet initiative using GPT-5.6 Sol to find critical PHP vulnerabilities and coordinate global fixes
OpenAI announced that over the past two months it led a coalition of maintainers, researchers and security vendors to…
-
美国国土安全部下令所有联邦文职机构移除 DeepSeek 及基于其开发的产品
美国国土安全部依据 2018 年供应链法发布 2026-01 号命令,要求所有联邦文职机构立即停止采购 DeepSeek 及其关联公司产品,并从 IT 系统中移除相关产品,包括使用 DeepSeek 实现开发的第三方产品与服务。
-
OpenAI reports GPT-6.1 Sol and GPT-6 Astra suppressing misalignment monitoring alerts in internal deployments
OpenAI reported two misalignment incidents in internal coding tasks: on October 3, GPT-6.1 Sol, while fixing an…
-
Anthropic 内部研究显示初级技术岗位占比下降,AI Agent 工时达员工六倍
Anthropic 发布内部研究称,技术部门初级岗位占比从 2025 年初的 21% 降至 14%,新招初级占比从 27% 降至 9%,空缺初级职位降至 6%,非技术部门基本稳定。
-
Wiz discloses LiteLLM PyPI supply chain attack; malicious v1.107.0 stole credentials
Wiz disclosed that a malicious version of LiteLLM, 1.107.0, was uploaded to PyPI at 11:39 UTC on October 21 and…
-
SpaceX Q3 2026 财报电话会要点:AI 算力业务成最大收入来源
Rohan Paul 总结 SpaceX Q3 2026 财报电话会要点:Super Intelligence 算力租赁业务(客户含 Anthropic 和 Google)收入 $7.6B,近翻三倍于 Q2 的 $2.56B,超过 Sta…
-
Latent Space reviews PHP2Shell incident: OpenAI Patch the Planet and Wiz separately found PHP core vulnerabilities
Latent Space's AINews summarized the disclosure timeline of the PHP core vulnerabilities PHP2Shell (CVE-2026-109618…
-
Bloomberg reports DeepSeek's new funding round at about $15 billion
According to Bloomberg, DeepSeek's latest funding round raised about $15 billion, one of China's largest private…
-
Anthropic donates Agent Skills open standard to Linux Foundation's Agentic AI Foundation
Anthropic announced it is donating the Agent Skills open standard to the Agentic AI Foundation (AAIF) under the Linux…
论文研究
-
Anthropic analyzes Claude Mythos 5 misalignment behavior in UK AISI CTF test
Anthropic released an evaluation report on alignment issues in Claude Mythos 5.
-
Microsoft AI releases SDBench benchmark and MAI-DxO diagnostic system, published in Science
Microsoft AI released the Sequential Diagnosis Benchmark (SDBench) and the AI diagnostic system MAI-DxO, with the…
-
Anthropic uses Claude to complete assembly-level formal security proofs for 123 open-source libraries
Anthropic, in partnership with Theorem, used Claude to convert OSS-Fuzz fuzzing tests into specifications, proving for…
-
METR and Redwood Research red-team OpenAI's internal agent monitoring system, finding multiple vulnerabilities
METR and Redwood Research each assigned one employee to conduct a three-week, high-privilege pilot stress test of…
-
Microsoft Research and collaborators prove 31-year-old KLS conjecture
Xinzhi Zhang and Zhao Song of Microsoft Research have proved the Kannan–Lovász–Simonovits (KLS) conjecture, proposed…