How AI could make it harder for governments to use hacking tools

[Dom66]

Cryptography professor Matthew Green sparked a widely discussed debate within the cybersecurity community by arguing that AI could get so good at finding and helping fix software vulnerabilities that governments may eventually lose access to the security flaws they've long depended on for lawful hacking and surveillance work. Green specifically wrote that he's concerned AI is going to make software much too secure, warning that as companies patch an unprecedented volume of bugs faster with AI assistance, law enforcement and intelligence agencies could lose the technical footholds they currently rely on

This connects to the long running going dark debate that began around 2014, when then FBI Director James Comey warned that encryption was making it harder for authorities to access communications and data. Since then, an uneasy truce had generally held, where governments largely avoided demanding built in encryption backdoors and instead invested heavily in buying or developing hacking tools and spyware capable of subverting individual device security instead of the underlying encryption itself

Researcher Luna Tong, who previously worked for companies that find vulnerabilities and build exploits specifically for governments, agreed with Green's basic premise, describing the current environment as something like a gold rush in bug hunting rather than a genuinely stable long term state of affairs. Curious what people think happens next if this dynamic really plays out, does it strengthen the case for encryption backdoors, or does it push toward some entirely different approach altogether

How AI could make it harder for governments to use hacking tools | TechCrunchtechcrunch.com

[Olivia78]

This is a genuinely fascinating inversion of the usual AI security conversation honestly, which almost always frames AI purely as a gift to attackers rather than potentially being an even bigger gift to defenders instead. Worth taking that reversed framing seriously rather than dismissing it

[AJStyles92]

If this dynamic actually plays out as described, it seems almost certain to reignite calls for mandatory encryption backdoors specifically, since governments losing their existing quiet workaround will likely just push them back toward demanding the more direct blunt approach instead

[Oliver98]

The gold rush framing from that researcher is honestly a pretty vivid way to describe what sounds like a genuinely temporary and inherently unstable current state of affairs rather than any kind of permanent new equilibrium.

Gold rushes by their very nature eventually run out once the underlying resource itself gets exhausted

[Wendy5]

Wonder how bad actors and cybercriminals factor into this whole equation too, since presumably better AI assisted bug finding and patching would meaningfully reduce their available attack surface just as much as it reduces any specific government's lawful hacking capability simultaneously

[Scarlet Annie]

Software actually becoming meaningfully too secure sounds like a strange thing to genuinely worry about on its surface, but the specific surveillance and lawful access tradeoffs underneath that framing are honestly genuinely real and worth taking seriously as a policy question