[ExI] Mathematicians Show That AI Protections Will Always be incomplete​

[BillK]

On Sun, 14 Dec 2025 at 12:02, John Clark via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> *"If fewer computational resources are dedicated to safety than to
> capability, then safety issues such as jailbreaks will always exist. Can we
> align a language model externally without understanding how they work
> inside? The answer to this question is a resounding NO."*
>
> *Mathematicians Show That AI Protections Will Always be incomplete *
> <https://www.quantamagazine.org/cryptographers-show-that-ai-protections-will-always-have-holes-20251210/?mc_cid=db3cb01235&mc_eid=1b0caa9e8c>
>
> *John K Clark*
> _______________________________________________
>


Gemini discussed this article and agreed, but with an added suggestion.
Full review here - <https://gemini.google.com/share/4d9b2d770aed>
BillK

Gemini 3 Pro Thinking -

The claim is *partially correct*, but it requires nuance. The article does
not prove that *all* forms of AI safety are impossible; rather, it proves
that a specific, widely used *method* of security is fundamentally flawed.
*Conclusion*

The article is correct in asserting that *cheap, bolt-on AI protections are
mathematically destined to fail.* The claim that "AI security is
impossible" is true in the context of the current "filter-based" paradigm.
True security will likely require a fundamental shift toward ensuring the
AI models themselves simply *do not want* to answer harmful prompts, rather
than relying on a digital babysitter to stop them.

-----------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20251214/5a6f36b0/attachment.htm>