Pre-Extracted Steering Vectors, Ready to Use
Seven steering vector sets for Qwen3-8B, Mistral-7B, and Gemma-2-9B are on Hugging Face. Change model behavior at inference time, no fine-tuning required.
Blog
Notes from the lab.
Technical deep-dives, research updates, and tutorials on AI agent reliability, evaluation, and verification.
Rotalabs Libraries Are Now Apache 2.0
We relicensed all twelve Rotalabs libraries from AGPL-3.0 to Apache-2.0. Red Queen stays AGPL-3.0 and gains an explicit commercial option.
Evolutionary Adversarial Testing for AI Systems with Red Queen
Red-teaming that evolves its own attacks. rotalabs-redqueen uses quality-diversity search to discover diverse jailbreaks across single-turn, multi-turn, and agentic/MCP surfaces - reproducibly, and as audit-ready evidence.
Opus 4.8 Got Better at Everything Except Resisting You
Claude Opus 4.8 beats its predecessor on nearly every capability benchmark. It's also somewhat less robust to prompt injection than Opus 4.7 - and Anthropic's own first live...
Field-Theoretic Memory for AI Agents
We treat agent memory as continuous fields governed by partial differential equations instead of discrete database entries. The result: +116% F1 on multi-session reasoning and >99.8% collective intelligence...
Shared Context for AI Agents: Introducing rotalabs-context
AI agents that can't share what they know make the same mistakes independently. We're releasing rotalabs-context - a context intelligence engine for ingesting, searching, and subscribing to shared...
Statistical Rigor in LLM Evaluation
Why most LLM benchmarks are doing evaluation wrong, and how to fix it with confidence intervals, significance tests, and effect sizes.
Steering Vectors: Runtime Behavior Control for LLMs
How to extract behavioral directions from language models and apply them at inference time. A practical guide to rotalabs-steer.
Agent-to-Agent Networks: Trust Dynamics and Attack Surfaces in Moltbook
A technical analysis of trust dynamics, emergent behaviors, and security vulnerabilities in Moltbook - the first large-scale agent-to-agent social network.
Introducing the Rotalabs Open Source Ecosystem
We're releasing 12 packages for AI trust, evaluation, and reliability. Available on PyPI and npm, all AGPL-3.0 licensed.
When AI Agents Don't Know What They Don't Know
The security conversation around AI agents is stuck on identity and permissions. The harder problem is whether an agent should take an action - whether its reasoning is...
Multi-Agent Coordination Without Centralized Control
The standard architecture for coordinating autonomous systems assumes a command node. That works until the link goes down. Here's how to build coordination that survives contested environments.
Runtime AI Control for Contested Environments
Prompting is not a control mechanism. When AI operates in kill chains and beyond reliable comms, 'the model usually follows instructions' isn't acceptable. Here's what I've learned about...
MCP is Infrastructure. Trust is the Missing Layer.
The Model Context Protocol connects AI agents to the world. Everyone's focused on securing the connections. But you can authenticate every MCP call and still have a compromised...
The CoT Blind Spot: Why Watching What Models Say Isn't Enough
Chain-of-thought monitoring was supposed to let us supervise AI reasoning. But new research shows models only faithfully report their reasoning 25-39% of the time. If we can't trust...
Memory Poisoning: The Attack Vector Nobody's Ready For
A single compromised agent poisoned 87% of downstream decisions in 4 hours. As AI agents gain persistent memory, attackers are finding ways to corrupt it. Here's what the...
Prompt Injection Is Still Unsolved: What the Latest Research Actually Shows
A joint paper from OpenAI, Anthropic, and DeepMind bypassed 12 published defenses. NAACL 2025 broke 8 more. Here's where we actually stand on prompt injection, and why activation-level...
The Multi-Agent Trust Problem: Why Your AI Agents Can't Verify Each Other
As AI agents collaborate on complex tasks, a critical question emerges: how does Agent A know Agent B isn't compromised, hallucinating, or colluding? Current approaches don't have good...
Detecting AI Sandbagging with Activation Probes
First empirical demonstration of activation-level sandbagging detection. Linear probes achieve 90-96% accuracy across Mistral, Gemma, and Qwen models. Sandbagging representations are model-specific, and steering can reduce sandbagging by...