Designing Drugs for Pathogens That Don't Exist Yet: Inside Red Queen Bio, OpenAI's Biodefense Bet

The strangest thing about Red Queen Bio’s pitch is its timeline. The company designs antibody drugs against pathogens that have never infected anyone — because those pathogens, in many cases, have not been engineered yet. Its adversary is hypothetical, its countermeasures are preemptive, and its founding premise is that both sides of that race are now accelerating on the same engine: frontier AI.

On September 28, the Wall Street Journal’s Georgia Wells published a profile of the startup — an OpenAI-backed AI biosecurity company with $36 million raised to design antibody drugs against novel pathogens, “including bioweapons that AI systems themselves could design.” The piece landed at the top of the day’s tech aggregators, and for good reason: it is one of the clearest windows yet into how the AI industry is starting to spend real money on the tail risks of its own technology.

Where the company came from

Red Queen Bio was spun out of HelixNano, a clinical-stage mRNA therapeutics company whose founders had spent nearly a decade using machine learning to compress drug development cycles. In November 2025, the spinout announced a $15 million seed round led by OpenAI itself — a rare instance of a frontier lab taking the lead position in a biosecurity startup. As part of the deal, OpenAI CEO Sam Altman and board member Nicole Seligman, who had previously invested in HelixNano personally, agreed to receive shares in the new company.

The name is a thesis. The Red Queen hypothesis, borrowed from evolutionary biology, describes species locked in co-evolution with ever-evolving opponents — the “now, here, you see, it takes all the running you can do, to keep in the same place” problem. The company’s stated mission is to “scale biological defenses faster than frontier AI capabilities grow.” Everything else in its pipeline follows from that framing.

What it actually builds

According to the company and its coverage, Red Queen Bio’s pipeline integrates frontier AI models, laboratory automation, and reinforcement learning with on-demand biologics manufacturing. The goal is an “AI-powered biodefense layer” that can design a countermeasure — a new antibody drug — almost instantly after a threat is identified, manufacture it at unprecedented speed anywhere it is needed, and deliver it to the people most exposed. The explicit ambition is to stop emerging threats before they become pandemics.

Two design constraints are worth highlighting, because they define the company’s safety posture:

• No gain-of-function research. The company states flatly that it never conducts dangerous gain-of-function experiments and never makes or isolates dangerous pathogens. Its threat models come from computation, not cultivation.
• Proprietary defensive datasets. The company generates its own data specifically to “asymmetrically accelerate defensive capabilities” — the bet being that the same AI techniques that lower the cost of engineering a pathogen also lower the cost of engineering the antidote, if you point them at the right problem.

The antibody work is not theoretical. In January 2026, AbTherx announced a therapeutic antibody discovery partnership with Red Queen Bio, under which the partners generate antibodies against targets Red Queen Bio selects. Initial candidates were already reported in development at the time of the announcement.

The OpenAI context

The seed investment did not happen in isolation. It sits inside a broader — and increasingly formalized — biosecurity push by OpenAI:

• Rosalind Biodefense. In May 2026, OpenAI launched Rosalind Biodefense, a program that expands trusted access to GPT-Rosalind, its life-sciences model, for vetted outside developers and U.S. government and allied partners supporting public health and biodefense missions.
• Valthos. OpenAI also put $30 million into Valthos, a startup building real-time biological threat detection — bringing its disclosed startup biosecurity investments alongside Red Queen Bio to roughly $45 million.

The strategic logic is straightforward, if uncomfortable: the companies racing to build increasingly capable AI systems are also among the few actors with both the incentive and the balance sheet to insure against those systems being misused. A WSJ profile of Red Queen Bio is, in that sense, also a story about OpenAI hedging its own existential bet.

Why this matters now

The timing of the profile is not incidental. Three currents in 2026 make “AI vs. AI-enabled biology” a mainstream story rather than a biosecurity-conference talking point:

The capability curve is real. Frontier models now routinely demonstrate wet-lab-relevant competence — protocol optimization, literature synthesis, experimental design. A widely cited collaboration between OpenAI and Red Queen Bio’s predecessor work reported that GPT-5, in controlled conditions, optimized a molecular cloning protocol. Defensive and offensive applications draw from the same capability pool.

Governments are lagging. As the New York Times argued in a feature published the day before this profile, AI deployment has outpaced regulatory capacity worldwide, and biosecurity is among the hardest areas to police because the underlying science is dual-use by default. Private biodefense is filling a vacuum that public institutions have not.

The funding thesis has shifted. A decade ago, pandemic preparedness was considered a philanthropic lane. Red Queen Bio raised its seed as a venture-backed public benefit corporation structured around “defensive co-scaling” — coupling defensive capability growth to the growth of the AI systems that create the risk. Whether that structure can actually keep pace is the open question, but investors are now underwriting the attempt.

The open questions

Skeptics have fair points. A designed antibody is years from being a deployable drug; the company’s countermeasures exist largely in silico and in early discovery partnerships. “Design almost instantly” solves only the first and cheapest stage of biodefense — manufacturing, distribution, and clinical validation remain the slow, expensive parts that no amount of inference speed fixes. And there is an inherent tension in the funder relationship: the same company that stands to benefit from AI’s acceleration is financing the insurance policy against it, which invites scrutiny of how independently the risk assessments are really governed.

But as a signal of where the industry is heading, the profile is hard to ignore. The AI sector’s first wave of safety spending went to model behavior — guardrails, evaluations, red-teaming. The second wave is going to the physical world: drugs, diagnostics, manufacturing. Red Queen Bio, with its $36 million, its OpenAI backing, and its Red Queen framing, is currently the most visible specimen of that second wave.

The race, as the company’s namesake would put it, has begun — and both sides are running.