President Trump and leaders of some of the nation's largest AI companies signed the White House Accord on Superintelligence on Tuesday, a voluntary agreement that calls on companies to strengthen internal controls around biosecurity risks.
The accord calls for internal monitoring, independent assessments, and board oversight. But it leaves companies largely responsible for determining what those safeguards should look like and when a risk has become serious enough to act.
Raj Panjabi, a former White House biosecurity official, said the key question is not just whether AI can write dangerous biological instructions, but whether it can move a design from the digital world into the physical one. He said that a model that can write a research protocol to create a pathogen is not a person who can create it.
Panjabi pointed to the need for synthesized DNA, equipment, and a laboratory with four walls to grow a virus. He said those barriers have not yet been dissolved by AI, but he argued that this is 'not a reason to let up.'
He said he is watching for evidence that someone without advanced biological expertise used AI to overcome experimental problems that previously required expert judgment. He also said he is looking for an AI-generated sequence of concern being ordered from a synthesis provider or run through an automated laboratory.
Panjabi cited warning signs that have already surfaced in public debate. Bill Gates has warned that artificial intelligence crossed thresholds this year, including bio-attack capability. Anthropic's chief executive urged the U.N. Security Council to back a ban on the use of AI to make biological weapons after the company disclosed five cases of biological misuse of its models.
Those cases included a researcher studying how H5N1 influenza adapts to mammals and causes illness beyond the respiratory tract. Anthropic banned the accounts. Panjabi noted that there is almost no population immunity to H5N1, roughly half of human cases are fatal, and the virus does not yet spread efficiently between people.
He compared the problem to the way crisis planning works in government. In his account, the hardest calls come before everyone agrees there is an emergency, which is why clear triggers matter. He argued that the AI industry has not yet applied that kind of trigger-action discipline to biology.
The market has already voted on the promise of AI. What is still missing is a serious private-sector standard for the downside risk. Voluntary promises are not the same thing as hard guardrails, especially when the stakes involve biosecurity, public health, and taxpayer exposure to the next emergency response.
