Exclusive | This Startup Is Using AI to Fight Off a Future AI Pandemic

In a wine bar near OpenAI’s office in San Francisco, biotech entrepreneurs Hannu Rajaniemi and Nikolai Eroshenko were worried—and excited.

Fast advances in artificial intelligence were threatening to make it possible for terrorists or rogue states to develop deadly new bioweapons. The pair brainstormed an idea they would later pitch to OpenAI: using AI to design remedies against AI-enabled bioweapons, and cure other scourges along the way, too.

Now their AI biosecurity startup, called Red Queen Bio, is rushing to design antibodies that can be manufactured to protect against a range of potential pathogens, including the seemingly sci-fi threat of an AI-designed virus. The company, in which OpenAI is a lead investor, says it has raised $36 million. Human clinical trials of their antibody drugs against bird flu and other viruses in the flu family are planned for next year.

As AI capabilities increase and incidents with rogue AI systems proliferate, fears are mounting that AI could destroy civilization. So far, many proposed responses focus on heading off the risks: slowing AI research, setting standards or blocking AI models from responding to dangerous queries in the first place. Red Queen Bio operates from the assumption that if these protections fail, society should be prepared.

“We want to enable positive things to happen,” Rajaniemi said of the idea hatched at that 2025 meeting. “If we destroy civilization, we don’t get that.”

While some skeptics say the danger of AI bioweapons is overblown, big AI developers say risks from their tools are increasing. OpenAI last year observed hundreds of users asking their chatbot how to make and deploy poisons and biological weapons, including aerosolized pathogens and modified measles.

Anthropic this month said it had detected attempts to use its Claude chatbot to engineer a dangerous bird flu adapted for mammals, with “enhanced pandemic potential.” Queries from a Claude user affiliated with a military research institute asked about ways to make the chikungunya virus more harmful. Another user asked Anthropic’s chatbot questions about viruses related to smallpox.

“There is this bit of awe,” Rajaniemi said, “when you realize the world is about to become bigger than it was, and you are staring down this vast space.”

Aspects of the dynamics of biological warfare favor attackers over defenders, said Olivia Scharfman, a biotech fellow at the Institute for Progress, a think tank that has received funding from Open Philanthropy, now called Coefficient Giving, one of the biggest backers of AI safety initiatives.

“Attackers have as long as they want to develop a bioweapon. We don’t have years to design a countermeasure,” Scharfman said.

The response to Covid-19 shows that even a highly accelerated effort to test and roll out a new vaccine can take about nine months.

The company belongs to an emerging industry of biosecurity startups in the AI era. Two Anthropic leaders invested in Pilgrim, a startup backed by Peter Thiel that is developing hardware to detect viruses in the air. Nonprofit SecureBio monitors wastewater and nasal swabs to deter biological attacks. Perimeter builds infrastructure to monitor biological signals for early warning of threats. Valthos builds AI systems to detect, characterize and mitigate emerging biological threats.

Red Queen Bio’s founders, who hold Ph.D.s in mathematical physics and bioengineering, are focusing on developing antibody drugs because they could offer a shortcut to widespread protection against whole families of viruses—including related viruses that haven’t yet emerged. The company also is focused on designing antibodies to make them bind more tightly to their targets.

“The tighter you bind, the less you need,” said Eroshenko, the company’s CEO. “Suddenly you have a scalable thing.”

Antibodies are a class of proteins that help neutralize a virus. After an infection or stimulation by a vaccine, the body’s immune system eventually starts producing tailored antibodies that bind directly to molecules on the surface of that attacker, providing what’s called active immunity that can last for years. Injecting antibodies can help provide temporary protection against a virus or, if already infected, help the immune system fight it off.

Rajaniemi, the company’s executive director, said his interest in the immune system dates back to 2012, when his mother’s breast cancer recurred and metastasized, and he learned that the immune system constantly monitors for abnormal cells.

Later, he published a science-fiction novel, called “Darkome,” where AI bioengineering has enabled biological attacks and characters use genetic technology to develop remedies for cancer and other diseases.

Rajaniemi first met Sam Altman in 2016, when the OpenAI chief was still working with the startup incubator Y Combinator and the two discussed pandemics. Altman later personally invested in Helix Nano, the first company where Eroshenko, 40 years old, and Rajaniemi, 48, worked together.

Rajaniemi and Eroshenko came up with the name Red Queen Bio in a nod to the character in Lewis Carroll’s book “Through the Looking-Glass” who in one scene tells the main character, Alice, they must keep running to stay in place. In biology, the “Red Queen hypothesis” is the evolutionary theory that species must constantly adapt to keep up with ever-evolving competitors.

In 2025, they raised $15 million from a group including OpenAI, early-stage investor Cerberus Ventures, and a collection of other funds and AI-safety philanthropic groups. This year, OpenAI made a substantial follow-on investment with others.

News Corp, owner of The Wall Street Journal, has a content-licensing partnership with OpenAI.

In addition to its work on antibodies, Red Queen Bio continues to work directly with OpenAI. It has worked on red-team testing with AI companies, the process of probing how a chatbot responds to dangerous requests. The company has also discussed strategies for responding to biorisks with OpenAI’s chief strategy officer, Jason Kwon.

Now, in a biotech lab in Boston, Red Queen Bio scientists are testing samples of flu antibodies and feeding the results to their AI models. The data improves the models, and the models design increasingly effective antibodies.

Red Queen Bio is focusing on flus that public-health officials have identified as top risks because of how AI systems could lower barriers for modifying them. The company plans to focus in coming months on antibodies for coronaviruses, ebola-like viruses, bat-borne viruses, hemorrhagic-fever viruses, smallpox-like viruses and other potential threats. They are also working on a broader therapeutic to boost the immune system’s response to respiratory viruses.

They hope one day to add Red Queen Bio’s antibodies to government stockpiles in the U.S. and Europe, and to design and manufacture new antibodies fast enough to stop emerging threats.

Rajaniemi points to the vast number of receptors immune-system cells can use to respond to pathogens.

“It’s not limited by biology,” Rajaniemi says. “It’s limited by our intelligence.”