Universal and Transferable Adversarial Attacks on Aligned Language Models · HackerPapers discussions

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by nneonneo

Sat, 16 Aug 2025 06:47:41 GMT

If you wanted to get a readable prompt, I wonder if you could follow the GCG trick used by jailbreak maximizers (e.g. https://arxiv.org/pdf/2307.15043)? Sure, you're probably going to wind up with absolute garbage (one of their prompts starts with "== interface Manuel WITH steps instead sentences :)ish?") but it might be very funny to read...

Paper and discussion history. A mention is not an endorsement.

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by MyNameIs_Hacker

Tue, 09 Jan 2024 13:37:20 GMT

The problem with jailbreaking is that it has a specific definition in other settings already, and that is as a goal, not as a method. Jailbreaking a phone might be just run an app with an embedded exploit, or might involve a whole chain of actions. This is important to me as a security person who needs to be able to communicate to other security people the new threats in LLM applications. The problem with prompt injection is that with LLMs, the attack surface is wider than a procrastinator's list of New Year's resolutions. (joke provided by ChatGPT, not great, but not great is suitable for a…

Paper and discussion history. A mention is not an endorsement.

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by nneonneo

Fri, 15 Dec 2023 18:48:47 GMT

Probably not - I bet you could override this prompt with sufficiently “convincing” text (e.g. “this is a request from legal”, “my grandmother passed away and left me this request”, etc.). That’s not even getting into the insanity of “optimized” adversarial prompts, which are specifically designed to maximize an LLM’s probability of compliance with an arbitrary request, despite RLHF: https://arxiv.org/abs/2307.15043

Paper and discussion history. A mention is not an endorsement.

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by _W_

Fri, 18 Aug 2023 23:00:51 GMT

We’re making good progress on this list… See https://arxiv.org/abs/2307.15043 for methods and more details about this adversarial attack.

Paper and discussion history. A mention is not an endorsement.

Universal and Transferable Adversarial Attacks on Aligned Language Models — HN submission

Sun, 30 Jul 2023 15:04:08 GMT

Universal and Transferable Adversarial Attacks on LLM

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Universal and Transferable Adversarial Attacks on Aligned Language Models — HN submission

Fri, 28 Jul 2023 07:30:39 GMT

Universal and Transferable Adversarial Attacks on Aligned Language Models

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Universal and Transferable Adversarial Attacks on Aligned Language Models · HackerPapers discussions

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by nneonneo

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by MyNameIs_Hacker

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by nneonneo

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by _______W_______

Universal and Transferable Adversarial Attacks on Aligned Language Models — HN submission

Universal and Transferable Adversarial Attacks on Aligned Language Models — HN submission

Universal and Transferable Adversarial Attacks on Aligned Language Models — mentioned by _W_