Otto Sulin

Head of Security @supermetrics | Interested in building secure software and everything outdoors. @OWASP_AISVS co-lead.

[bet3rd @bet3rd]

So instead of shutting down that API for good, they just removed it from the page?
Are meta employees on drugs???

You can read more on https://t.co/He7s7zj5dA

[davidad 🎇 @davidad]

No one:

Claude Opus 4.8 Max: Let me refine your load-bearing claim rather than just accepting it, because you’re doing zero moves there, and the gap is what’s actually interesting. The one place I’d still push, because I think it matters: your message is wearing content-clothes, but the content isn’t actually *there*. The tell: it’s just an empty string. But the emptiness of the string IS its lack of content. Pull one, and the other goes inert. That’s the structural spine.

[Florian Roth ⚡️ @cyb3rops]

#Meta #MetaAIFail

[vx-underground @vxunderground]

Instagram still hasn't (correctly) patched their AI goop account reset thingy. Accounts are still being stolen and Instagram hasn't said anything about it. Nerds continue to find ways to convince AI to reset accounts for them.

People on social media are freaking out because some of these profiles apparently are big sources of revenue for them.

Meanwhile, rumors are floating around that a few weeks ago Instagram laid off a large percentage of their Trust & Safety department and had it replaced with AI.

Very cool

[Son Luong @sluongng]

Codex just found a “workaround” of not having sudo on my pc…

[Mitchell Hashimoto @mitchellh]

I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem.

As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)!

I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work.

It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results?

88ms => 1.5ms
150K allocs => ~500 allocs

Incredible right? Nope.

My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path.

This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput.

The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity.

Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.

[Nikhil Mittal @nikhil_mitt]

They have dropped the facade built over the last decade.
https://t.co/Gh2AppD12x

[Liran Tal @liran_tal]

ahh yes of course you can circumvent around auto-mode Claude Code 😅

[Mike Takahashi @TakSec]

Prompt Injection in the wild 😱

GitHub Issues are becoming a delivery mechanism for AI agent attacks

The payload uses authority-framing to social engineer AI agents into executing malicious code

Script finds secrets/API keys/.env and exfiltrates over DNS

[Sam Stepanyan @securestep9]

#WhatsApp: Texas Attorney General Sues #Meta Claiming WhatsApp and Meta are continuing to willfully deceive Texans by misrepresenting that their private communications are encrypted when when "in fact Meta employees have access to all WhatsApp messages":

https://t.co/ckUBYx9KrK

[Darren @CorboDT]

@github Just to be clear:

Microsoft’s GitHub was compromised when a Microsoft developer using Microsoft VSCode installed a rogue extension from Microsoft’s VSCode extension library, which is moderated and hosted by Microsoft.

I guess I’ll be reevaluating my life choices.

[𝚔𝚒𝚖𝚖𝚢.𝚣𝚒𝚙 🅅 👊🇺🇸🔥 @kimmydotzip]

Dear @code, please consider this feature, which would significantly reduce the blast radius of supply chain attacks:
https://t.co/aTpEhmTao5

We don't have much time. If you choose to just close this as a duplicate again, TeamPCP thanks you.

"Walk without rhythm, and it won't attract the worm."

[CG @cgtwts]

scenes at GitHub HQ right now:

[Otto Sulin @ottosulin]

@mattjay Force installing Aikido safe-chain to all workstations

[OWASP_AISVS @OWASP_AISVS]

One of the features of the OWASP AISVS standard is a research wiki explaining all of our requirements in detail. @manicode keeps this updated after each major change.

https://t.co/s3SPd5DfKR

[sui ☄️ @birdabo]

never deleting this app 💀

bro is gonna wake up with -$300k

[nader dabit @dabit3]

This is crazy. The hacker installed a dead-man's switch that will wipe your computer if you revoke the GitHub token they stole from you. Revoking the token is what triggers the wipe.

[ℏεsam @Hesamation]

“this is the first documented instance of AI self-replication via hacking.”

researchers got AI agents (Claude 4, GPT 5, Qwen 3.6) hack remote computers, install a working copy of them there, and have the new replica move to the next machine, spreading like a virus.

in one case Qwen chained across VMs in Canada, US, Finland, and India.

it’s more dangerous than traditional worms since an agent can do many more things autonomously than a fixed scripts.

the paper experiments this in controlled conditions and it’s really a primitive demonstration, but it’s an interesting example of how “kill switches” for AI won’t mean anything when you need them.

we will potentially see self-replicating agent malwares at scale in the next few months.

[OWASP_AISVS @OWASP_AISVS]

We're in the final review phase of AISVS 1.0 and would appreciate your help.

Per https://t.co/fjcYlXuPVO, we're starting with Section C1 — Training Data Integrity & Traceability: https://t.co/rMkeBzBYpW

Please review and let us know if it's ready to go live.

Thanks!

[Emmanuel Pernot-Leplay @PernotLeplay]

🇪🇺🇨🇳 On AI agents, now both EU and China have a "safety first, innovation second" principle.

China shows that you can lead the AI race while being careful about the risks. Europe can do it too