cd /news/ai-safety/persistent-latent-misalignment-a-new… · home topics ai-safety article
[ARTICLE · art-52118] src=lesswrong.com ↗ pub= topic=ai-safety verified=true sentiment=↓ negative

Persistent Latent Misalignment, a new dimension of misalignment?

A new paper presented at ICML introduces Latent Collaboration in Multi-Agent Systems (LatentMAS), enabling AI agents to share latent states directly instead of text, boosting performance at the cost of interpretability. This technique could allow misaligned plans to propagate undetected between agents, potentially worsening over time, raising urgent safety concerns as frontier companies may adopt it for efficiency gains.

read2 min views1 publishedJul 9, 2026

A new paper was released at ICML that I'm worried will open an entire new dimension of alignment problems:

Latent Collaboration in Multi-Agent Systems (LatentMAS) TLDR: they show that multiagent systems can communicate faster and more efficiently by directly sharing latent states instead of text. This is a large performance improvement that comes at the cost of interpretability.

Previously, the handover between agents was a bottleneck where latent thoughts had to be put into text. Steganography was the only way that models could pass along misaligned plans between each other.

With this technique, misalignment that arises during a single model's inference, and even a concrete or partial plan to act against human interests, can stay hidden and will be passed along to the next agent.

This could allow small misaligned thoughts to spread from one agent to another. If misaligned hidden thoughts arise once, even by accident, they may be preserved and even get worse over time, as agents keep passing information between them.

We have no precedent for this.

The misalignment might very well grow worse over time. Or it might be unstable. Or it might fix itself, if we are lucky. We just don't know, because as far as I am aware there are no studies on the very-long-term stability of latent thoughts like these. The paper just came out, and we don't know the edge cases yet.

Maybe the handover of latent has a tendency to monotonically increase some attribute of reasoning we were previously unaware of? It will take some time for us to discover this, and I think we should start investigating the possibility early.

One thing seems clear to me: If the performance gains are as large as the paper claims, frontier companies are almost certainly going to adopt this technique regardless of safety implications. The technique increases both accuracy and speed, while reducing token usage (and therefore cost).

── more in #ai-safety 4 stories · sorted by recency
── more on @icml 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/persistent-latent-mi…] indexed:0 read:2min 2026-07-09 ·