SECURITY

New paper and code on sequential poisoning attacks in AI

By

–

05 June 2026 15h24

Lots more in the paper: how does DPO fit into the picture? What if attackers have different goals? etc. Paper: https://
arxiv.org/abs/2606.04929
Code: https://
github.com/jcksanderson/s
equential-poisoning
… Led by @jcksanderson
, w/ @YihanWww
, Xiaoqian Lu, co-supervised w/ @YiweiLu3r 6/6

0.5% poison breaks reward model, 5% needed for RLHF transfer

By

–

05 June 2026 15h24

What about poisoning PPO? A remarkable paper of @javirandor and @florian_tramer (
https://
arxiv.org/abs/2311.14455) shows that just 0.5% poison is enough to break a reward model (L)! Again, fear not: somehow, it takes a (high) 5% poisoning before it transfers to the RLHF'd model (R). 4/n

2% SFT poisoning gives 90% attack success; RLHF wipes it away

By

–

05 June 2026 15h24

There's multiple post-training phases attackers can infiltrate: SFT, DPO, PPO. Let's start with SFT. With just 2% SFT poisoning, 90% attack success (L)! But not to worry, RLHF works as we hope (?): it wipes away the poison. An RM scores outputs just like a clean model (R). 3/n

LLM post-training pipelines vulnerable to combined data poisoning attacks

By

–

05 June 2026 15h24

Feeling safe against data poisoning in post-training? Think again! Individual components of LLM post-training pipelines are surprisingly robust to data poisoning attacks. In work led by @jcksanderson (co-advised w @YiweiLu3r
), we show they crumble when attacked together. 1/n

Backdoor attacks on LLMs via untrusted training data

By

–

05 June 2026 15h24

LLMs are trained on lots of data, often from untrusted sources. This is particularly true in safety post-training, where data is gathered from human responses. Attackers can try to sneak in a backdoor: if there's a trigger in the prompt, bypass safety guardrails. 2/n

→ View original post on X — @kimmonismus

LTX Studio and LTX-2.3 Bridge Idea and Use Gap

By

@kimmonismus

–

05 June 2026 14h57

1/ Most AI video tools still feel like demos.
You type a prompt → you get a clip.

But the real bottleneck was never generation.
It was turning an idea into something usable.

With LTX Studio + LTX-2.3, that gap is basically collapsing.

The clips I just made felt… different.… pic.twitter.com/GZcBEJ1Lcw
— Chubby♨️ (@kimmonismus) 5 juin 2026

1/ Most AI video tools still feel like demos. You type a prompt → you get a clip. But the real bottleneck was never generation. It was turning an idea into something usable. With LTX Studio + LTX-2.3, that gap is basically collapsing. The clips I just made felt… different.

→ View original post on X — @aihighlight

Mythos wrote 181 Firefox exploits in testing, leading to gating.

By

@aihighlight

–

05 June 2026 14h11

Mythos was never held back over timing. It wrote working Firefox exploits 181 times in testing. That's the reason it's gated.

→ View original post on X — @ceobillionaire

QUEBEC.AI launches Sovereign AI: Capability under control, building, governing, securing, benefiting.

By

@ceobillionaire

–

05 June 2026 13h18

http://
QUEBEC.AI launches Sovereign AI: AI‑First sovereignty = capability under control. Data, infrastructure, agents, proof, governance. Quebec enters AI‑First sovereignty: build, govern, secure, benefit. http://
quebecartificialintelligence.com/sovereign-ai #AIFirst #QuebecAI