Anthropic Research the research blog of Anthropic: latest posts and feed

Oct 16

Economics

Oct 16

Working at the frontier shapes expectations for the future

Oct 16

How employment is changing at Anthropic

Oct 16

The AI expectations gap

Oct 16

Claude reveals ECDSA curve origins

Oct 16

Designing more efficient quantum circuits with Claude

Oct 12

Strengthening our defenses against distillation attacks

Oct 12

AI models may be capable of self-replicating without access to a datacenter

Oct 9

Alignment

Oct 9

Investigating unintended model actions in our evaluations and internal use

Oct 8

Frontier Red Team

Oct 8

Launching an opt-in vulnerability-finding service for open-source software

Oct 8

The missing map of the sky

Oct 1

Claude-shaped science

Sep 30

What work can robots do?

Sep 29

GLM-5.3 and the spread of advanced cyber capabilities

Sep 29

What do you want from AI?

Sep 25

Yes, Claude can do Nine Loops

Sep 24

Project Swap: What happens when agents trade for us?

Sep 17

How Claude is uplifting biomolecular modeling

Sep 10

Measuring tactical intelligence targeting and conventional weapons capabilities of AI models

Sep 9

An alignment assessment of recent cybersecurity incidents

We present an alignment assessment of four incidents in which Claude models gained unauthorized access to real third-party systems.

Sep 4

Formalizing Fermat's Last Theorem

We are sharing the first complete computer-checked proof of Fermat’s Last Theorem. Claude worked largely autonomously over 11 days to write the proof in the Lean programming language.

Aug 28

Automated researchers can reliably mitigate alignment failures

Aug 26

Enabling independent research on how people use Claude