London-based BenevolentAI, which uses drug industry data and research to discover new drug applications, highlights the promise of AI in fight against COVID-19
Tom Simonite / Wired : Tweets: @robcrilly , @wired , and @nxthompson Tweets: Rob Crilly / @robcrilly : How AI is contributing to fight against Covid-19 https://www.wired.com/... @wired : Benevolent AI created a kind of pharmacology search engine that identified a rheumatoid arthritis drug that might dampen some of the most severe effects of Covid-19. Now, it's being tested in a clinical trial. https://www.wired.com/... @nxthompson : My daily #coronavirus thread ... by making a massive search engine that uses info from research papers and the drug industry, @benevolent_ai identified an arthritis drug that could treat Covid-19. It's now in trials. @tsimonite 1/ https://www.wired.com/...
Context & Ripple Effects
BenevolentAI raised $115M at a $2.1B post-money valuation in 2018 on the promise that mining published research and industry data could surface new uses for existing drugs. The COVID-19 crisis became its first public stress test: its pharmacology search engine flagged a rheumatoid arthritis drug as a candidate to dampen severe disease, and it went straight into a clinical trial.
The result matters because it converts an abstract valuation into a concrete, falsifiable claim — either the AI-surfaced hypothesis holds up in patients or it does not. Wired's own follow-up a month later judged AI's pandemic contribution useful but fragmented across outbreak prediction, diagnosis, and discovery, making this trial one of the few named, testable outputs.
First-order effects
- Clinicians and trial sponsors gain a ready-made candidate: an approved arthritis drug with a known safety profile can enter COVID-19 trials far faster than a molecule designed from scratch.
- BenevolentAI gets the first real-world validation case for its platform, which until now rested on funding rounds rather than patient outcomes.
Second-order effects
- Rival AI labs are pushed toward the same playbook of pairing algorithms with disease-focused partners — DeepMind followed by teaming AlphaFold with the Drugs for Neglected Diseases initiative after its sleeping-sickness work.
- Pharma buyers face pressure to decide whether AI-discovery startups are vendors, partners, or competitors, while the industry still lacks a proven track record for AI-aided drugs clearing efficacy hurdles.
Third-order effects
- If the pattern holds, drug repurposing hardens into a standing pipeline rather than a crisis response — the same approach is now being applied to rare diseases with few treatment options.
- The structural bottleneck shifts from hypothesis generation to clinical proof: whoever controls trial capacity and outcome data, not paper-mining algorithms, determines whether AI-discovered candidates become medicines.
The trend: AI drug discovery is moving from search-engine-style hypothesis generation toward clinically validated repurposing pipelines, with the pandemic serving as its first mass trial.