82搜索为您找到约 1,513 条结果
www.anthropic.com
网页
综合
An alignment assessment of recent cybersecurity incidents \ Anthropic
…entified a fourth incident, from January 2026, involving an early version of Claude Opus 4.6. We have notified all affected parties. After finding this incident, we broadened our search to r…
www.anthropic.com
网页
综合
Introducing Agent Skills | Claude by Anthropic
… endpoint gives developers programmatic control over custom skill versioning and management. Skills require the Code Execution Tool beta, which provides the secure environment they need to r…
www.anthropic.com
网页
综合
Prompting Claude's long context window \ Anthropic
…uation, we primarily focused on our smaller Claude Instant model (version 1.2) as opposed to our Claude 2 model. This is for a couple of reasons. First, Claude 2 is already very good at reca…
www.anthropic.com
网页
综合
Mapping AI-enabled cyber threats \ Anthropic
…es (or TTPs) described in those summaries, and mapped them to the version of the MITRE ATT&CK framework that was live at that time (V18). In all, we observed 13,873 actions across 482 unique…
www.anthropic.com
网页
综合
Testing and mitigating elections-related risks \ Anthropic
…itial PVT on election administration topics found that an earlier version of Claude would sometimes provide incorrect or outdated information in response to questions about voter registratio…
www.anthropic.com
网页
综合
Previewing the Model Hardware Standard \ Anthropic
…from hardware errors without intervention. We’re sharing an early version of MHS with partners across science, robotics, electronics, and manufacturing so we can collaborate to build safety …
www.anthropic.com
网页
综合
Our updated Responsible Scaling Policy \ Anthropic
…entation and learning, we are now sharing a significantly updated version that reflects practical insights and accounts for advancing technological capabilities. Although this policy focuses…
www.anthropic.com
网页
综合
Measuring LLMs’ ability to develop exploits \ Anthropic
… Preview’s performance, and also ran Mythos Preview on an updated version of SCONE-bench, a benchmark we developed in collaboration with MATS and the Anthropic Fellows Program to measure sma…
www.anthropic.com
网页
综合
Claude does cyber competitions \ Anthropic
…. Part of why we could achieve such speed is that we had multiple versions of Claude running at the same time tackling different challenges. But scaling up AI agents is arguably easier than …
www.anthropic.com
网页
综合
Investigating reward tampering in language models \ Anthropic
… model to engage in reward tampering: we gave a model access to a version of its own code, and left open an opportunity for it to edit the code such that it gained a much higher reward for c…
第 146 / 152 页