anthropic.com · Research

Research listings from anthropic.com on TechiSeek. 43 public listings.

• LLM-discovered 0 days \ Anthropic (anthropic.com)
  AI models can now find high-severity vulnerabilities at scale. We're using Claude to find and help fix vulnerabilities in open source softw…
• Moral self-correction in large language models \ Anthropic (anthropic.com)
  RLHF-trained models can avoid stereotyped or biased outputs when instructed to, a capability that emerges around 22B parameters.
• Introducing the Anthropic Economic Index \ Anthropic (anthropic.com)
  Announcement of the new Anthropic Economic Index and description of the new data on AI use in occupations
• Softmax linear units \ Anthropic (anthropic.com)
  Replacing standard MLP activations with SoLU substantially increases the fraction of interpretable neurons with no performance loss.
• A small number of samples can poison LLMs \ Anthropic (anthropic.com)
  Anthropic research on data poisoning attacks shows a few hundred malicious documents can "backdoor" large language models of any size.
• Scaling laws of learning from repeated data \ Anthropic (anthropic.com)
  Repeating just 0.1% of training data 100 times made an 800M-parameter model perform as poorly as a 400M one.
• Finding bugs with Claude and property-based testing \ Anthropic (anthropic.com)
  We developed a Claude-powered agent that efficiently identifies bugs in large software projects across the Python ecosystem.
• Mitigating prompt injections in browser use \ Anthropic (anthropic.com)
  Claude Opus 4.5 achieves a 1% attack success rate against prompt injections in browser use, and Claude for Chrome moves to beta.
• Project Pilot: Can AI models fly drones? \ Anthropic (anthropic.com)
  We worked with Andon Labs on Drone-Bench, a new benchmark testing whether AI models can autonomously fly a drone to locate and follow a per…
• Project Fetch: Can Claude train a robot dog? \ Anthropic (anthropic.com)
  A practical experiment on AI's ability to affect the physical world
• Measuring progress on scalable oversight \ Anthropic (anthropic.com)
  Humans working with an unreliable LLM assistant outperformed both the model alone and unaided humans on challenging question answering.
• Measuring AI agent autonomy in practice \ Anthropic (anthropic.com)
  Analysis of millions of human-agent interactions shows agent turns getting longer, with software engineering making up nearly half of tool…
• Emergent introspective awareness in LLMs \ Anthropic (anthropic.com)
  Research on the ability of large language models to introspect on their own internal states.
• Labor market impacts of AI: A new measure \ Anthropic (anthropic.com)
  We introduce "observed exposure," combining LLM capability with real Claude usage data, and share early evidence on employment effects.
• India Country Brief: Anthropic Economic Index \ Anthropic (anthropic.com)
  India is Claude's second-largest market globally. Findings on where and how Indian users adopt Claude, from software development to complex…
• How Canada uses Claude \ Anthropic (anthropic.com)
  Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
• Project Glasswing: An initial update \ Anthropic (anthropic.com)
  An early update on what we've learned from Project Glasswing.
• How Australia uses Claude \ Anthropic (anthropic.com)
  Australia uses Claude at 4.1x the expected rate per capita, with broader professional diversity than the global average.
• How AI is transforming work at Anthropic \ Anthropic (anthropic.com)
  An inside look at how Anthropic employees use Claude in their daily work, and what that reveals about AI's impact on the workplace.
• Evaluating Claude with BioMysteryBench \ Anthropic (anthropic.com)
  BioMysteryBench tests Claude on 99 real bioinformatics problems. Opus 4.6 solved 77% of human-solvable questions and some that experts coul…
• Evaluating feature steering: A case study in mitigating social biases \ Anthropic (anthropic.com)
  A new piece of Anthropic research by Durmus et al.: "Evaluating feature steering: A case study in mitigating social biases"
• Evaluating and Mitigating Discrimination in Language Model Decisions (anthropic.com)
  Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
• Economic Index: New building blocks for AI use \ Anthropic (anthropic.com)
  Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
• Claude Opus 4 and 4.1 can now end a rare subset of conversations \ Anthropic (anthropic.com)
  An update on our exploratory research on model welfare
• Emotion concepts in a large language model \ Anthropic (anthropic.com)
  All modern language models sometimes act like they have emotions. What’s behind these behaviors? Our interpretability team investigates.
• Announcing the Anthropic Economic Index Survey \ Anthropic (anthropic.com)
  A monthly study gathering real-time data from Claude users on AI's effect on work, hiring, and productivity, complementing traditional labo…
• Anthropic Economic Index report: Cadences \ Anthropic (anthropic.com)
  The latest Economic Index report looks at when people come to Claude, what they produce with it, and how they perceive AI's impact on their…
• Anthropic Economic Index report: Learning curves \ Anthropic (anthropic.com)
  The March 2026 Anthropic Economic Index report finds Claude use cases diversifying and experienced users seeing higher conversation success…
• Commitments on model deprecation and preservation \ Anthropic (anthropic.com)
  Anthropic commits to preserving the weights of all publicly released Claude models and interviewing retiring models about their preferences.
• Collective Constitutional AI \ Anthropic (anthropic.com)
  Anthropic and the Collective Intelligence Project had ~1,000 Americans draft an AI constitution, then trained a model on it.
• How Claude Code is used in practice \ Anthropic (anthropic.com)
  New Anthropic research looking at interactive agentic coding. We evaluate the composition of tasks, human-AI collaboration, and success rat…
• Cyber evaluations of Claude 4 \ Anthropic (anthropic.com)
  We partnered with Pattern Labs on a range of cybersecurity evaluations of Claude Opus 4 and Claude Sonnet 4, with Opus demonstrating especi…
• attack navigator (anthropic.com)
  ResearchPolicyCommitmentsLearnNewsTry ClaudeFrontier Red TeamMapping AI-enabled cyber threats: Insights from the LLM ATT&CK NavigatorJun 3,…
• Introducing Anthropic Interviewer \ Anthropic (anthropic.com)
  What 1,250 professionals told us about working with AI
• Focus areas for The Anthropic Institute \ Anthropic (anthropic.com)
  The Anthropic Institute investigates AI's impact on the world from within a frontier lab. Here are the questions driving its agenda.
• Economic Index: Uneven AI adoption \ Anthropic (anthropic.com)
  We extend the Anthropic Economic Index with a geographic analysis of Claude.ai conversations and a first look at enterprise API use.
• Anthropic Economic Index: Insights from Claude 3.7 Sonnet \ Anthropic (anthropic.com)
  The second update from the Anthropic Economic Index
• Education Report: How educators use Claude \ Anthropic (anthropic.com)
  Research on 74,000 educator conversations shows how faculty use Claude for teaching, research, and building interactive learning tools.
• Economic Index report: Economic primitives \ Anthropic (anthropic.com)
  New metrics of AI usage provide a rich portrait of interactions with Claude in November 2025, just prior to the release of Opus 4.5.
• What 81,000 people told us about AI economics \ Anthropic (anthropic.com)
  Anthropic's recent survey of 81,000 Claude users provides a way to connect people’s economic concerns with what we’ve quantified in Claude…
• Anthropic’s response to Governor Newsom’s AI working group draft report \ Anthropic (anthropic.com)
  Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
• Anthropic Education Report: How University Students Use Claude \ Anthropic (anthropic.com)
  AI systems are no longer just specialized research tools: they’re everyday academic companions. As AIs integrate more deeply into education…
• The Situation Report \ Anthropic (anthropic.com)
  A rare Ebola strain is spreading in the DRC, with no confirmed vaccine. Global health organizations are using Claude to respond as fast as…