Dario Amodei says AI could make cancer, Alzheimer’s ‘relics of the past’ in less than a decade as Anthropic expands AI-driven drug discovery push

Anthropic is ramping up its work in the area of using AI in the discovery of new drugs, partnering with six companies from the biotech and pharmaceutical sectors to study how its Claude models can help in the development of medicine for cancer, food allergies, and rare diseases.

The collaborations include BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines, and The Column Group. Along with these collaborations, Anthropic has also launched its own drug discovery programs, including studies of rare childhood versions of muscular dystrophy.

"You know, if today we look at things like cancer, Alzheimer's, heart disease, mental illness, these will be relics of the past in, I think, less than a decade," the leading AI firm's CEO Dario Amodei said.


Also Read | Anthropic retracts details of $69 billion IPO filing sent to bankers. Why analysts are raising worries over valuations

"I think if we do this right, we can create an age of miracles where humanity can solve problems that have eluded us for millennia," he said.

As stated by the organisation, such efforts have enabled it to discover some ways that AI technology can be used at various levels of medicine development, ranging from molecular design to clinical trials.

Also Read | Anthropic’s IPO is said to be multiple times oversubscribed

“There is no single problem to solve,” Anthropic said in a research update, explaining that challenges are spread across the drug development process. The company said it intends to increase its involvement in the field as it learns more about the capabilities and limitations of its models.

AI-assisted research produces early results in cancer and allergy programmes

One of the collaborations between Generate Biomedicines and Claude involved the redesign of instructions in the CAR T-cell therapy. This is a process whereby the patient's cells are modified to detect and fight cancer.

Claude Code was used to redesign the instructions in the CAR-T cell by a team of six scientists.

The new instructions had an almost 70% shorter signalling domain compared to the typical instructions, and it produced about ten times the number of receptors on the surfaces of human T-cells.

Further, the scientists observed that these new instructions were able to kill cancerous cells much quicker, and they lasted longer. Generate and Anthropic are currently exploring the potential of using this technique for killing solid tumours with specific markings on cancer cells.

Another project, involving BigHat Biosciences, entailed optimisation of two antibody fragments, which could potentially be used in the development of drugs. Around 1,010 different variants were optimised by Claude over about 10 weeks, based on suggestions and analysis of laboratory data.

There was noticeable improvement in the strength and stability of the most efficient candidates. Some of them were 10 times more binding than the original compounds, whereas one cancer-specific fragment stayed stable at a temperature 18 degrees Celsius higher.

Claude has been put to use by Divergence Labs in order to investigate the possible development of a novel therapy for food allergy. Scientists at Divergence Labs are engaged in developing antibodies that are capable of eliminating IgE or immunoglobulin E from immune cells as opposed to just neutralising IgE in the bloodstream.

Claude designed roughly 35,000 antibodies within one month of the commencement of research. Laboratory tests revealed that there were three antibodies that were capable of binding to IgE and 20 that were capable of removing IgE from the immune cells. Of these, three were more active than the regular treatment drug, according to the company.

The next step would be combining the two components and determining if it would be able to remove the bound IgE without provoking an allergic reaction.

AI moves beyond molecule design to clinical planning and drug development operations

Additionally, Anthropic’s work goes beyond the task of developing medicines.

At Enveda, Claude was employed to examine manufacturing and regulatory documents related to the production of pharmaceuticals. Errors in the examination of these documents could slow down the development of drugs.

The company said Claude has gone through 1,061 pages of documents and identified all 11 problems pointed out by the human reviewers and seven additional problems. The document examination time was cut by 89% from more than 25 hours to less than three hours. Claude was also utilised for preparing regulatory documents.

The collaboration between Bristol Myers Squibb and Anthropic involves the use of Claude to analyse clinical trials during the early stages. The company is examining the capability of Claude in analysing the plan of the clinical trial and its chances of success.

Michael Pourdehnad, Senior Vice President & Head of Early Clinical Development, Hematology, Oncology & Cell Therapy at Bristol Myers Squibb, said that the initial analysis made by Claude proved helpful in analysing different development options.

He said: “From a single prompt, Claude generated a robust initial analysis.”

He intends to take the partnership to the next level in 2027 to see if the technology could help in making decisions regarding drug development at its earliest stages.

The Column Group, on the other hand, has launched a new company called Pleiades with a plan to deploy Claude in a wide array of drug discovery programs.

The project is meant to cut down on the scientific, operational, and administrative efforts required in the management of biotech projects. The tool which was built through the project has already undergone tests on 45 target-disease pairs.

Anthropic, on the other hand, has acquired a company called Mithril Medicine, which is a one-man company focused on the development of drugs to treat muscular dystrophy. The founder of the firm, Cody Tranbarger, used Claude in coordinating contracts, experiments and work done by 11 external labs.

The company claims that one of its potential drugs moved from conception to mouse trials in less than 18 months. Anthropic anticipates that a drug could be ready for human trials in late 2028 if more promising results are achieved in further studies.

Human trials remain a significant milestone

Despite these initial results, Anthropic stated that most of the science is still preclinical.

This means that the drug candidates are not ready to undergo clinical testing, and laboratory findings do not guarantee the safety and effectiveness of the treatment for patients.

Anthropic expressed an intention to apply the expertise of Claude to clinical research and collaborate with even more organisations working on developing drugs.

Besides, the company strives to make its solutions easily accessible to the scientific community, for instance, by launching Claude Science, life scientist credits and the Life Sciences Verification Programme.

A call for guest researchers who want to use Claude to solve their problems in drug discovery or other spheres of life sciences has been announced by the company.

These new initiatives demonstrate the willingness of the company to go beyond the stage of molecular screening.

Nevertheless, whether these tools will be able to help in developing the medicine will become known only after additional laboratory studies and preclinical trials.