이미지: Anthropic
Summary
- Anthropic has launched a Claude Team plan for scientists worldwide, opening 10,000 seats at free or discounted rates for a year.
- The AI for Science credit program is expanding from a biology focus to all scientific fields, offering up to $50,000 per project.
- Biology and chemistry researchers can, for now, only access models up to the Opus tier, with Mythos-level access opening through a separate program built jointly with the U.S. government.
Claude arrives at the lab door
On August 27, 2026, local time, Anthropic announced a Claude Team plan for scientists that opens Claude subscriptions to researchers worldwide for free or at a discount. The program starts with 10,000 seats: standard seats are free for a year, while premium seats with five times the usage cap cost $15 a month. Anthropic said it plans to expand the program well beyond that initial 10,000 seats over the coming months, framing the free access not as a one-time promotion but as an ongoing program it intends to keep growing.
To qualify, applicants need to be a principal investigator (PI) or equivalent researcher at an academic institution or nonprofit research organization. Once verified through an application form, a PI can add lab members to the same plan. Rather than reaching individual researchers one at a time, the structure hands Claude access to entire labs at once.
The two seat tiers differ mainly in usage capacity. Standard seats come with the same limits as a regular subscription and are free, while premium seats carry five times that cap, aimed at researchers running simulation code or data analysis around the clock. Either way, the term is one year, and for now that deal covers 10,000 seats.
A subscription like a scholarship, credits like a grant
The announcement goes beyond subscriptions. Anthropic is also expanding the scope and scale of its AI for Science program, which grants free credits to high-impact scientific projects. Until now that program mainly supported biological science research; it's now opening up to other scientific fields as well. Anthropic pointed to compute-heavy, ambitious research — like progress on the Riemann zeta function or Claude's work on protein design — as the kind of project it wants to support.
Labs that exhaust their subscription limits can apply to AI for Science for up to $50,000 in credits per project. Unlike the subscription plan, credit applications aren't limited to PIs — any researcher can apply. It resembles how universities cover tuition through scholarships while funding lab materials separately through research grants: one pocket for steady monthly usage, another for the occasional heavy computational push.
Claude Science, which launched last June, is another piece of this picture. It's a research-focused product that consolidates tools researchers commonly use, helps produce results others can verify, and gives flexible access to computing resources. Anthropic said that as Claude grows more capable at scientific research, the company isn't just selling models — it's building products and programs to support the research community around them.
Put together, the support comes in three layers: subscriptions for everyday use, credits for large-scale computational projects, and Claude Science for the working environment itself. Anthropic said the goal is to raise how much scientists actually use Claude, and in doing so, radically accelerate scientific discovery.
Not every door is open
Despite the expansive announcement, not everything is unlocked. Biology and chemistry researchers will, for now, only be able to access models up to the Opus tier. Anthropic said its newest model, Claude Fable, will continue to block specialized biology and drug-development queries because of dual-use risk — the concern that knowledge useful for legitimate research could also aid weapons development.
Instead, Anthropic is building a separate path around that restriction. The company said it's working with the U.S. government on an access program that would let life-sciences experts use Mythos-tier models for research and development, and that it has already enrolled its first participants. Details and any broader rollout, it said, will be shared soon. In effect, the same announcement that widens access also tightens control over the most sensitive layer.
Applications begin with the verification form on Anthropic's announcement page. Once a PI's eligibility is confirmed, they can add lab members to the plan. Projects that need credits can apply separately through the AI for Science program.
Editor's take
The real story here isn't the discount — it's the distribution channel. By requiring PI verification and bundling access at the lab level, Anthropic is positioning Claude not as a tool for individual researchers but as standard infrastructure for the lab itself. A grad student who first learns a tool through a lab account rarely switches away from it after graduating. Those 10,000 seats look less like a revenue play and more like an investment in the working habits of the next generation of scientists.
At the same time, the announcement signals that access to frontier capability is starting to look more like a licensing system. Within the same model family, general research stays open, sensitive areas in biology and chemistry are capped at the Opus tier, and anything beyond that is only available through a separate program built with government involvement — a three-tier structure. It's worth watching how the competition among frontier AI labs to support science is starting to hinge less on price and more on this kind of access design: who gets how much, and under what terms.
For researchers in Korea, there are two things worth checking right away: whether your institution meets the academic or nonprofit eligibility criteria, and whether the application needs to come from a PI. Since credits are open to any researcher, projects with heavy computational needs might be worth pursuing through AI for Science independently of the subscription plan.
Comments