AI Could Reshape the Balance Between Encryption and Government Hacking
Cryptographer Matthew Green warns that AI may make software so secure that governments lose the ability to lawfully hack devices, sparking debate among cybersecurity experts.
Cryptography professor Matthew Green sparked debate in the cybersecurity community this August with a viral thread and blog post raising a provocative question: could artificial intelligence make software so secure that law enforcement and intelligence agencies lose their ability to lawfully hack into criminals' devices?
Green argued that AI could help companies fix an unprecedented number of software bugs, potentially cutting off the flow of vulnerabilities that governments currently rely on for surveillance. This echoes the "going dark" debate popularized in 2014, when then-FBI Director James Comey warned that encryption was making it harder for authorities to monitor suspects. Since then, apps like Signal, WhatsApp, and Apple's iMessage introduced end-to-end encryption, and device makers like Apple began encrypting data by default.
According to Green, governments have so far managed by purchasing hacking tools and spyware rather than demanding backdoors be built into devices. He fears AI's growing ability to rapidly find security flaws at scale could upend this arrangement, potentially reviving government demands for backdoor access.
Experts split on the outcome
Researcher Luna Tong, who has worked at firms developing exploits for governments, agreed with Green, describing a current "gold rush of bugs" that she expects to be temporary. Paolo Stagno, CTO of Crowdfense, a company that buys and sells zero-day vulnerabilities to governments, said no state will give up surveillance capabilities, and called the current system requiring governments to exploit found flaws the "most democratic system we have" — though he warned it may not last.
Others pushed back. Hamid Kashfi, founder of DarkCell, said that for every AI-found bug that gets reported, roughly 20 may go unreported by researchers who prefer to keep them secret. Eva Galperin of the Electronic Frontier Foundation noted that "vibe-coding" with AI tools is actually introducing more vulnerabilities, though she cautioned that finding more bugs doesn't guarantee they get patched quickly.
Katie Moussouris, founder of Luta Security, said there is still "some distance to go" before modern devices become largely bug-free. She suggested pressure for backdoors could intensify eventually, but likely not before the next U.S. presidential election.
