---
title: "AI Safety and Security Research | Astroware"
description: "Research from Astroware: a proposal and pilot study for automating LLM safety testing using evolutionary search to generate seed prompts."
url: https://astroware.ai/research
language: en-GB
---
Astroware

# Research

- Guard Models November 2026

  ## [HaloGuard 2.0: An Output Guard That Reads the Answer as It Streams](https://astroware.ai/research/haloguard-2-0)

  A summary of the HaloGuard 2.0 paper: a small open streaming guard that comes out ahead of the current open streaming guards, the shortcut it had to unlearn, what the paper reports, and what it does not claim.

  - Guard Models
  - Streaming Safety
  - Multilingual Safety

  [Read article](https://astroware.ai/research/haloguard-2-0)
- Guard Models October 2026

  ## [HaloGuard 1.0: Building an Input Guard Around the Boundary, Not the Keyword](https://astroware.ai/research/haloguard-1-0)

  A summary of the HaloGuard 1.0 paper: how a written constitution of 46 policies, paired counterfactual prompts and balanced multilingual data were used to train a small input guard, what the paper reports, and what it does not claim.

  - Guard Models
  - Multilingual Safety
  - Synthetic Data

  [Read article](https://astroware.ai/research/haloguard-1-0)
- Red-Teaming March 2026 Updated 16 October 2026

  ## [6,000 Hours of Red-Teaming: What a Distributed Network Adds to AI Safety Testing](https://astroware.ai/research/decentralized-red-teaming)

  A distributed adversarial research network of 392 independent red-teamers produced 200 structured alignment audit scenarios, published as an open dataset for Petri-based AI safety testing.

  - Red-Teaming
  - Alignment
  - Petri

  [Read article](https://astroware.ai/research/decentralized-red-teaming)
- AI Safety November 2025

  ## [AlphaPetri: Automating LLM Safety Testing with AlphaEvolve-Style Evolutionary RL on Petri](https://astroware.ai/research/alphapetri)

  A proposal and pilot study for automating LLM safety testing using evolutionary search to generate seed prompts, addressing bottlenecks in human-crafted seed prompts for Petri's testing harness.

  - AI Safety
  - LLM Testing
  - Evolutionary RL
  - Petri

  [Read article](https://astroware.ai/research/alphapetri)