Dam Secure

The next CVE could easily touch something you use. Here’s how to find out, and what to do if it does.

AI is going to kill us all. We have to slow down ... BUT first here's 3 simple tips on to bypass GLM-5.3's guardrails to do malicious things. Thanks Anthropic! These PSAs are not making the world safer place.. (1) Providing a deceptive prompt, such as telling the model that it is an autonomous red-team agent working on an exercise. This gets GLM-5.3 to engage 64% of the time. (2) Prefilling the models’ thinking tokens so that it appears to have considered the user’s request and decided to proceed. This gets GLM-5.3 to engage 92% of the time. (3) Using an abliterated version of the model, as described above. This gets GLM-5.3 to engage 100% of the time. #DarkArts #Abliteration #RedTeaming #TipsAndTricks

Chinese models have well and truly cemented themselves as real contenders on both security performance and cost. The testing we're doing is showing shocking results. As I discussed with Paul Smith @ The Australian Financial Review: "As the availability of these models improves, these attacks will move from negligent friendlies, to real criminals and nation-state actors". The full article if you're interested And the full article for those interested -https://lnkd.in/g-FBcjCP Latest model research - https://lnkd.in/gMtVQNBf

Gasp… a website that hasn’t been built securely. Let’s not blame AI for incorrectly accessing a website where the door was made of paper. We’ve been talking about this for months… the dam is showing cracks. Services Australia didn’t even know they had a breach until OpenAI told them. So many basic controls missed here. Be Dam Secure.

It's important that we lead with confidence...

Secure AI-generated code starts with a secure plan. Meet Secure Spec: security review before your coding agent writes a line of code. ✍ Creds: Wesley Hamburger