Dimi Trizna

AI & Security | ex-Microsoft | Founding MTS @weareaisle | Agentic AI and adversarial ML

[Dimi Trizna @ditrizna]

@MSFTBlueHat organized awesome event for the first time coming in India — was happy to be part of it! Recording of the talk is coming ✌️

[Microsoft BlueHat @MSFTBlueHat]

Dmitrijs Trizna (@ditrizna), Senior Security Researcher at Microsoft, is presenting his #BlueHatIndia talk: “The Impact of Backdoor Poisoning Vulnerabilities on AI-Based Threat Detectors.” In his talk, Dmitrijs discussed AI-based defenses: threat model: living-off-the-land, data augmentation, and machine learning and attacks on AI models: poisoning vulnerabilities, backdoor intuition, and results. Dmitrijs shared a few take-home messages at the end of his talk:
✅AI/ML provide methods to improve classical defenses.
✅Introduction of AI/ML brings new attack vectors. Think about security of AI in your solutions.
✅If you are a defense engineer, be aware of AI/ML risks.
✅As a red teamer: Explore novel attack vectors.

[Orange Tsai 🍊 @orange_8361]

That's my chain — a full chain w/ logic bugs only! No memory corruption, no AI, and of course no collisions at all 😉

[Dimi Trizna @ditrizna]

@GitHubSecurity Tbh, it’s ridiculous having only $2.3k paid in bounties given such work by the community 🤦

[Stanislav Fort @stanislavfort]

AISLE has discovered 20 of 23 OpenSSL zero-days (CVEs) across the last 3 consecutive security releases

Latest release: 5 of 7 are AISLE
1 was co-reported by Anthropic (Mythos?) 63 days after AISLE

OpenSSL encrypts 2/3 of the internet

10 fixes accepted straight into production

[Stanislav Fort @stanislavfort]

New post: We show that small, cheap models can detect the flagship Mythos FreeBSD zero-day (CVE-2026-4747) using a simple harness we call nano-analyzer

Models down to 3.6B active params (including open-weights ones you can run locally) would have detected it 100-1000x cheaper

[clem 🤗 @ClementDelangue]

"But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases in their announcement, isolated the relevant code, and ran them through small, cheap, open-weights models. Those models recovered much of the same analysis. Eight out of eight models detected Mythos's flagship FreeBSD exploit, including one with only 3.6 billion active parameters costing $0.11 per million tokens. A 5.1B-active open model recovered the core chain of the 27-year-old OpenBSD bug." https://t.co/yBTiiMq1Xy

[klöss @kloss_xyz]

let me explain the importance of this

an engineer solved a problem that’s been plaguing the Internet for 3 decades

every website you’ve ever used relies on a text layout system from the 1990s

the browser loads a font, measures text, figures out where lines break, and positions everything vertically

every step depends on the previous one… every step forces the browser to pause and recalculate

you’ve felt this problem plenty times before even if you didn’t know what caused it:

→ Slack’s scroll jumping when message heights are wrong

→ Google Docs getting slow on long documents because every keystroke recalculates everything below your cursor

→ AI chat apps getting janky when streaming because each new token can cause a line wrap that shifts the entire page

same root cause every damn time.

text measurement is locked inside the browser’s DOM… it’s slow… and there’s been no alternative… for 30 damn years

Pretext bypasses all of it:

→ pure TypeScript text measurement… no DOM… no CSS… no browser reflow

→ you give it text, a font, and a width... it returns exact line breaks, widths, and heights… using pure math

→ around 500x faster in many cases than the standard approach

→ supports every language including mixed bidirectional text, CJK, Japanese, Korean, Arabic, and emojis

→ the engine is 15 kilobytes

→ built and validated by running Claude Code and Codex against browser ground truth for weeks

the demos are wild:

→ hundreds of thousands of text boxes virtualized at 120fps with no DOM measurement

→ shrinkwrapped chat bubbles with zero wasted pixels… something CSS literally cannot do

→ responsive multi-column magazine layouts that reflow dynamically

→ variable font ASCII art

over the years, developers moved rendering to Canvas… scrolling to custom implementations… positioning to JS

but text was the one thing you couldn’t move out of the browser… it was the last piece locked inside the DOM with no alternative

now we have a solution

this was built by Cheng Lou… one of the foundational developers behind React, Facebook Messenger, and Midjourney.

he’s not just anyone… lol

if you build anything on the web, this now changes what’s literally possible

this unlocks new UI patterns, layouts, interfaces, and experiences like we’ve never seen before

go look at the demos in the quote posts

it’s open source.

npm install @chenglou/pretext

insane these are all running in a browser​​​​​​

the future of design is still to come

[Dimi Trizna @ditrizna]

@IceSolst @anton_chuvakin now consider how many cases are **not** accidentally discovered?

[Stanislav Fort @stanislavfort]

AISLE is now the #1 source of accepted security findings in OpenClaw, the fastest-growing AI agent framework. Our AI discovered 15 vulnerabilities: 1 Critical (CVSS 9.4), 9 High, 5 Moderate. 21% of all OpenClaw security advisories globally are from us, more than anyone else ⏬

[Jeff Dean @JeffDean]

Cool application of AI models to find security vulnerabilities. Nice work, @stanislavfort!

[Dimi Trizna @ditrizna]

@rharang Thanks Rich! We need more calibrated, non-black/white opinions vocalized.

LLM technology is amazing, and it’s proper integration will yield unbelievable results, and if some don’t see it — it’s a skill issue. But please, people on both ends of spectrum, be realistic

[Dimi Trizna @ditrizna]

@moo_hax “melting scaffolds” is nice emphasis—we repeatedly come to observations that simplest architectures work the best with latest generations of models

it’s interesting to witness what was needed for some pipeline to work 10 months ago, and how good the thing is out of the box now

[Dimi Trizna @ditrizna]

@moo_hax @rharang @Microsoft quite unexpected

pre-LLM conventional cybersec skillset screams to get out

feeling the same quite often :)

[Dimi Trizna @ditrizna]

@giffmana Agree, this was interesting moment of catching myself not stereotyping on Jurgen, and be open minded to what he shares.

[Dimi Trizna @ditrizna]

@ClementDelangue @ylecun unbelievably immature argument from such a visible person

I’m no doomer, actually quite techno-optimistic, but this level of argumentation is raising radicalism and division, reducing productive discourse

please read “turkey example” from Taleb’s Black Swan

[Stanislav Fort @stanislavfort]

At AISLE @WeAreAisle we've surfaced & reported a ❗critical severity❗ vulnerability in Samba with the "perfect" 10.0 / 10.0 CVSS rating.

✨ CVE-2025-10230 ✨

Hidden for 13 years in production code. Samba is central to Windows/Linux cross-platform infra.

Blog post below

[Stanislav Fort @stanislavfort]

Another CVE detected by Aisle's AI system in the world's critical software infrastructure!

This time in cURL which has over 10B installations across devices & applications.

There aren't many more higher impact projects than this! Super proud of our team at @WeAreAisle 🔥🔥🔥

[Dimi Trizna @ditrizna]

@jiqizhixin everyone reading AI 2027 got a red flag alert

[Dimi Trizna @ditrizna]

@IceSolst @CyberQueenMara defines where problem lies:

> All of this points to an uncomfortable truth: We automated attacks against ourselves without the benefit of automated remediation or mitigation.

imho, researchers should send reports only w/ validated fix suggestion

[Dimi Trizna @ditrizna]

@vxunderground that’s how black hat hacking should look

[Dimi Trizna @ditrizna]

@vxunderground thanks for summarizing

imho, this is everlasting debate of security’s place in product development

the less mature environment, the less emphasis on cyber

AI is just a catalyst here, not something essential to the problem