Anthropic releases Claude Science for automating research

The release comes less than two weeks after the startup disabled access to its most advanced AI models due to a U.S. order to keep the technology out of the hands of foreign nationals. Photo by Nicolas TUCAT/AFP via Getty Images

Article content

Anthropic PBC is releasing new software aimed at helping scientists automate research, in the hopes of reducing some of the tedious aspects of their work.

Article content

Claude Science, which the company plans to roll out Tuesday, can be used to automate a range of biology and chemistry tasks, such as predicting protein structures, Anthropic said Tuesday in a statement. The software brings together a number of tools scientists commonly use, including more than 60 scientific databases, the firm said. The idea, according to the company, is to make it simpler for scientists to automate multistep tasks, ask questions using plain language, and get answers without having to query many individual sources of information.

Article content

Article content

Article content

Claude Science is being offered as a beta version to the company’s paid users, Anthropic said.

Article content

Article content

Anthropic and rival OpenAI have spent much of the past year developing AI tools to streamline a wider range of professional tasks — from financial services to science and healthcare — with the goal of courting more business customers and justifying their lofty valuations. Anthropic, now valued at US$965 billion, is pushing toward an initial public offering as soon as this fall.

Article content

Anthropic’s moves, in particular, have rattled markets in recent months, reflecting broader concerns about which companies and services will eventually be rendered obsolete by AI. In February, for instance, Anthropic introduced a tool for Claude Cowork to automate certain legal work, such as contract reviewing and legal briefings; that move helped trigger a US$1 trillion stock market rout.

Article content

Anthropic expects to announce Claude Science at an event in San Francisco where chief executive Dario Amodei will speak, along with executives including Chris Boerner, CEO of drug maker Bristol-Myers Squibb Co., and Vas Narasimhan, chief executive of drug maker Novartis AG and an Anthropic board member.

Article content

Article content

Anthropic said Claude Science uses existing widely available Claude models, such as Opus 4.8, which Anthropic released in May. The outputs from Claude will include traceable details that let scientists confirm the accuracy of the information, the company said. Images the app generates will include details of how it made them as well.

Article content

Article content

The release comes less than two weeks after the startup disabled access to its most advanced AI models — Fable 5 and Mythos 5 — due to a Trump administration order to keep the technology out of the hands of foreign nationals. On Friday, Anthropic won United States approval to restore some access to Mythos 5, after resolving Trump administration concerns about the technology’s potential threats to national security; there has been no announcement regarding any change to restrictions on use of the Fable 5 model.

Article content

Article content

Article content