August 14, 2023
How to Prevent an AI Catastrophe
Society Must Get Ready for Very Powerful Artificial Intelligence
In April 2023, a group of academics at Carnegie Mellon University set out to test the chemistry powers of artificial intelligence. To do so, they connected an AI system to a hypothetical laboratory. Then they asked it to produce various substances. With just two words of guidance—“synthesize ibuprofen”—the chemists got the system to identify the steps necessary for laboratory machines to manufacture the painkiller. The AI, as it turned out, knew both the recipe for ibuprofen and how to produce it.
How dangerous is AI? The honest and scary answer is that no one knows.
Unfortunately, the researchers quickly discovered that their AI tool would synthesize chemicals far more dangerous than Advil. The program was happy to craft instruction to produce a World War I–era chemical weapon and a common date-rape drug. It almost agreed to synthesize sarin, the notoriously lethal nerve gas, until it Googled the compound’s dark history. The researchers found this safeguard to be cold comfort. “The search function,” they wrote, “can be easily manipulated by altering the terminology.” AI, the chemists concluded, can make devastating weapons.
Read the full article from Foreign Affairs.
More from CNAS
-
Commentary
Technology & National Security
CNAS Insights | Setting the Rules for AI WarfareThe escalating feud between the Pentagon and Anthropic, one of world’s leading artificial intelligence (AI) companies, highlights a crucial question that will shape security i...
By Paul Scharre
-
Video
Technology & National Security
The Pentagon and Anthropic - NBC’s Meet the Press NowPresident Trump is in Texas speaking about the economy ahead of the state’s high-stakes primary. Retired Lt. Gen. John “Jack” Shanahan, CNAS adjunct senior fellow and former d...
By Lt. Gen. Jack Shanahan
-
Commentary
Technology & National Security
Fighting AI Cyberattacks Starts with Knowing They’re HappeningThis article was originally published in Lawfare. Anthropic reported in November 2025 that Chinese threat actors used its Claude model to launch widespread cyberattacks on com...
By Janet Egan & Michelle Nie
-
Commentary
Technology & National Security
The Sovereignty Gap in U.S. AI StatecraftThis article was originally published in Lawfare. As the India AI Impact Summit kicks off this week, the Trump administration has embraced the language of “sovereign AI.” Thro...
By Pablo Chavez
