The same Big Tech companies racing to build the most powerful AI systems in history are now quietly admitting those systems could help someone cook up a plague that makes COVID-19 look like a mild inconvenience.
A new report out of RAND is raising alarms that should have every American paying attention.
And the details buried inside that report paint a picture of an industry-wide threat that no single government, company, or safeguard is currently equipped to stop.
What the RAND Report Actually Says
RAND published a report that discussed several mitigating strategies for restricting bad actors from misusing AI technology to create biological weapons. The timing alone tells a story. This comes as Anthropic’s Claude built proteins that locked onto 14 of 15 targets in a lab setting, according to the company’s own announcement.
“AI-enabled biotechnology might lower the technical, operational, and motivational barriers to creating biological weapons,” the report warned. “It might also enable devastating attacks involving enhanced or novel pathogens, including events far worse than historical pandemics that caused many millions of fatalities globally.”
That is not fringe speculation from a fringe source. RAND is one of the most established research institutions in the country, and the language in that warning is as direct as it gets.
The report focuses on “high-consequence AI-enabled biological incidents, which we define as biological events capable of producing impacts on the order of ten times the scale of the coronavirus disease 2019 (COVID-19) pandemic or greater, societal disruption, or systemic destabilization,” according to the Purpose section of the report.
Ten times the scale of COVID-19. Let that land for a moment. COVID killed millions, locked down the entire world, and handed governments an excuse to strip civil liberties from their citizens for years. Whatever comes next, according to RAND, could be exponentially worse — and AI is the accelerant.
“No single safeguard can prevent a bad actor from using AI to help build a biological weapon,” RAND wrote in a recent X post. “But our new report identifies nine interventions that, layered together, could meaningfully lower the risk of a significant AI-enabled attack.” RAND lead author Steph Guerra believes the “most dangerous thresholds have not yet been crossed,” she told Axios.
Not yet. That qualifier is doing a lot of work in that sentence.
Still, Guerra argued that this unregulated technology is a threat and requires “mutually reinforcing approaches.” The report lays out nine specific interventions that, stacked on top of each other, could reduce the risk — but none of them alone is sufficient.
Anthropic’s Own Admission About What Its AI Can Do
Here is where it gets harder to dismiss as abstract policy concern.
Anthropic acknowledged in its own announcement that “without robust safety measures, they could enable bad actors to perform dangerous research, such as the development of bioweapons.” As a result, “protein design and other dual-use research biology capabilities remain unavailable for general access in Claude Fable 5.”
Read that again. Anthropic — one of the companies building the most powerful AI systems on the planet — is telling the world that its own flagship model is capable enough in biology that the company has decided to block general public access to those capabilities. That is not a hypothetical risk. That is a company drawing a line because it is genuinely worried about what happens if it does not.
Some of the technology enabling drug discovery work could turn dangerous in the wrong hands, Guerra told Axios. The same tools that might one day speed up cancer treatment could, in the hands of a bad actor, be pointed in a very different direction.
And the industry knows it. AI leaders at Google, OpenAI, and Anthropic previously urged Congress to pass a law requiring DNA and RNA sellers to screen buyers and keep proper records of their orders, the Wall Street Journal reported. These are the same companies that spent years insisting their technology was safe, beneficial, and self-regulating. Now they are asking Congress to step in and mandate biosecurity checks on the raw materials that could be used to build a bioweapon.
That is not a confident industry. That is an industry that has scared itself.
The Bigger Problem Nobody Wants to Say Out Loud
Anthropic CEO Dario Amodei has already warned, separately, that AI could wipe out half of all entry-level white-collar jobs and drive unemployment to 10 to 20 percent within one to five years. The same man is now running a company whose AI is capable enough in biology that his own team decided to restrict public access to those capabilities over bioweapons concerns.
Both of those things are true at the same time. The technology that threatens American workers’ livelihoods is also the technology that could, in the wrong hands, threaten American lives on a scale that dwarfs anything in living memory.
Different threat actors require different prevention mechanisms. Resource-constrained individuals might be thwarted at material or information access chokepoints. However, technologically sophisticated or well-resourced groups, such as state actors, might be immune to access controls and therefore require deterrence through elevated perceived cost.
That distinction matters enormously. Communist China, which systematically obscured the origins of COVID-19 and has been documented attempting to acquire American biotechnology through every available channel, is exactly the kind of state actor RAND is describing. A regime with the resources and the motivation to weaponize AI-assisted biology is not going to be stopped by a terms-of-service agreement or a voluntary screening program.
And the open-source problem is real. Open-source biological tools and data are more concerning — including those emerging from academia and research institutes. “Right now, data for biology is freely out there and open because biology… is for discovery and saving lives. It’s not warfare first, which is a good thing,” Guerra said. But that openness, which exists for entirely legitimate scientific reasons, is also the vulnerability that a bad actor could exploit.
The RAND report’s nine-intervention strategy is a serious attempt to address a serious problem. But nine interventions require nine separate institutions, agencies, companies, and governments to actually implement them — and to do so consistently, simultaneously, and without gaps that a sophisticated adversary can find and exploit. The history of pandemic preparedness, biosecurity funding, and interagency coordination in Washington, DC does not inspire confidence that any of that will happen quickly or well.
What this report really confirms is something Anthropic CEO Dario Amodei has been saying for years: the same technology being sold to the public as a productivity tool and a medical breakthrough engine is also, right now, capable of providing meaningful assistance to someone trying to build a weapon capable of killing millions of people.
But the industry keeps expanding. The data centers keep going up, driving up electricity bills for ordinary Americans while concentrating still more capability in the hands of a handful of Big Tech companies. The same companies that censored conservatives over COVID policy, questions about the 2020 election, and the events of January 6 are now the gatekeepers of technology that could, if misused, cause devastation that makes COVID look manageable.
That is an uncomfortable amount of power to trust to companies with that track record. And RAND just confirmed that the stakes of getting it wrong are not abstract. They are catastrophic.
Sources: Daily Caller News Foundation; RAND Corporation, “Building a Defense-in-Depth Biosecurity Strategy for the AI Era,” 2026; Axios; Anthropic research announcement; Wall Street Journal
