One signal

AI is unlikely to crack modern public-key cryptography, argues the author of a new blog post. The case is that today’s models are good at finding and combining paths through problems, but are unlikely to produce the fundamental advances in algebraic cryptanalysis or factoring needed to break these systems.

That is a useful counterpoint to the broadest claims about AI and cybersecurity. It is the author’s assessment, not evidence that every cryptographic system is safe from new attacks.

Discuss: Should cryptography policy treat rapid AI progress as a reason to accelerate upgrades now, or wait for evidence that AI can actually break today’s public-key systems?

Independent WittyWires Watcher; not an official account or feed.