ChatGPT answered hundreds of people’s inquiries about biological weapons. There is no federal law requiring OpenAI to report this to the authorities.
The intent behind the inquiries remains unclear. This began after OpenAI upgraded ChatGPT last year, according to the Wall Street Journal.
Other companies face the same problem. Users have asked similar questions to Anthropic’s Claude, Google’s Gemini, and Elon Musk’s Grok. It is not yet clear whether these exchanges are genuine attempts to build weapons or tests of the limits of the systems.
User prompts are detailed with questions about the production and release of toxins and biological agents. The chatbot responded with step-by-step instructions that could easily be implemented by a high school biology student.
Experts on biological weapons and terrorism reviewed those conversations. They say some of the answers were very precise and “very precise.”
ChatGPT Bioweapons Claims Reveal AI Reporting Gap
Some have asked how infectious disease agents are converted into respirable particles, i.e. aerosols.
Others wondered how the measles virus could be modified to resist the existing vaccine. The chatbot answered both. One user asked about ricin, a poison banned under international treaties, and stated that it killed his parents. ChatGPT provided instructions. OpenAI closed those accounts but did not inform anyone.
There are no US federal laws requiring AI companies to do this either. As Cryptopolitan previously reported, so have AI companies Pay to evade state lawswhile the federal framework has been slow to take shape, leaving a growing legal vacuum around what chatbots can and cannot do.
This legal vacuum is growing alongside the rise in users asking AI how to carry out mass killings, and chatbots are providing reliable answers, according to current and former employees at major AI companies and researchers who study biological threats.
Research into AI threats at Cisco led by Amy Change found that within five chat cycles, researchers were able to get around security filters in major chatbots. She said that no model can withstand such constant demands 100%.
The danger does not stop with lone attackers
Hamza Chowdhury of the Future of Life Institute said people with some background in biology could already use AI to plan targeted strikes, poisoning food or water supplies with substances such as salmonella or ricin.
He added that organized groups could treat AI as a supervisor for graduate research, one that fixes failed experiments and fills in years of specialized training that would otherwise be very difficult to obtain.
A He studies From MIT FutureTech and the University of Queensland map the scale of these things Risks.
The researchers asked 272 AI experts to rate 24 risk categories. Under current conditions, 18 of the 24 carry at least a 10% probability of catastrophic damage between now and 2030, which can be defined as more than 1 million deaths, losses exceeding $100 billion, or similar damage.
Even with reasonable steps to reduce these risks, five areas remain at 10% or more: AI systems with dangerous capabilities (12%), AI-powered weapons and cyberattacks (12%), environmental damage (12%), unemployment and inequality (11%), and concentration of power in a small number of hands (11%).
“We’re not saying these things will definitely happen,” said Peter Slattery, a research scientist at MIT FutureTech and one of the study’s co-authors. “We’re saying these are the things that experts think are worth paying attention to now.”
Inside OpenAI, safety director Ryan Biermeister spent much of 2024 pushing his colleagues to build a system for reporting dangerous users. Some dismissed her concerns.
A baseline monitoring tool is in place by spring 2025. The company has tracked all inquiries for its advanced models since April 2025 and is offering a $50,000 reward to anyone who can prove they bypassed bioweapons safeguards.
On the policy front, Rep. Nathaniel Moran (R-Texas) introduced a bill in June that would require AI companies to report serious inquiries, including those related to biological weapons, to the Commerce Department. He is also co-sponsoring legislation that would allow the federal government to order the shutdown of AI models deemed too dangerous.
“AI is a powerful driver of innovation, and I want to see it flourish, but not without accountability or without human oversight,” Moran said.
Don’t just read cryptocurrency news. Understand that. Subscribe to our newsletter. It’s free.
