Mathew Solnik retweeted
I'm not sure open letters are an effective tool, but in this case we need every tool we can get to raise awareness of the point we've reached: we have neglected defense and accelerated offense with AI. All the security concepts and patterns the industry has used and adopted over the years in product security are becoming wrong assumptions. One pattern every vendor follows is raising the bar for an attacker with new mitigations or new layers of obscurity to win time, letting complexity become a security boundary.
These old patterns are broken. The data shows they were never very effective against experienced attackers, and now we have an experienced attacker that can be scaled with unprecedented velocity, so winning time is a losing strategy by design. Defenders today don't have time, but we still have an opportunity to change the rules of the game.
Mathew Solnik retweeted
Seems like a really smart time for the US Government to kneecap the American leader in this benchmark! 👏👏👏
Exciting news: GLM-5.2 (Max) ranks #2 in Code Arena: Frontend, with +29pt over Claude Opus 4.7 (Thinking) and only behind Fable 5! GLM-5.2 is the best open model vs Kimi-K2.6 and Minimax-M3 by a large margin.
- #2 React and #4 HTML sub-leaderboards
- Ranks as the top model in nearly all sub categories: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Gaming, and Simulations.
Congrats @Zai_org for the incredible milestone!
Mathew Solnik retweeted
exploitbench.ai/
Awesome work from Seunghyun and Prof. Brumley!
Mathew Solnik retweeted
In light of the OpenSSH RCE advisory published today by Qualys: blog.qualys.com/vulnerabilit…, where it references our March blog while discussing ASLR weaknesses being key to feasible modern i386 exploitation, here it is again in case you missed it:
A weakness 23 years in the making: binaries and libraries built with an older toolchain act as timebombs against ASLR under "recent" Linux kernel and glibc changes.
Users: Check your exposure!
Developers: Rebuild binaries to achieve full ASLR benefit!
grsecurity.net/toolchain_nec…
Mathew Solnik retweeted
CVE-2024-6387 - Signal Handling Race condition results in remote root exploit of OpenSSH w/glibc. Found by @qualys - tagging @mdowd as this looks familiar to his prior SSH work and is referenced. qualys.com/2024/07/01/cve-20…
Mathew Solnik retweeted
Whoa.. still not convinced of AI Agents? This might change your mind...
I pretended to be a fake shoe company and gave AutoGPT a simple objective:
- Do market research for waterproof shoes
- Get the top 5 competitors and give me a report of their pros & cons
Here's how it went:












