Table of Contents
Chutes is integrating Halo Guard Alpha, an AI safety system developed by Astroware Labs and its Bittensor subnet, Trishool (SN23), to help secure user interactions across its flagship products.
This is not your typical subnet partners with another subnet announcement.
— Trishool | SN23 (@trishoolai) May 18, 2026
It's a mainstream AI lab adopting the product of another AI lab building on BitTensor
📢 @chutes_ai is now a customer of Trishool.
Chutes will be deploying Halo as the security layer for its two… pic.twitter.com/jR9GAshheZ
The deployment will bring Halo Guard Alpha into Chutes Chat and Fictio, where the system will function as a security layer designed to detect jailbreak attempts, prompt exploits, and other forms of adversarial behavior that can compromise AI applications.
Halo Guard is trained through a live adversarial process on Trishool’s subnet, where miners are incentivized to attack and break the model. Successful exploits are fed back into training, allowing the system to continuously improve against emerging attack methods.
In a post announcing the integration, Chutes said Halo currently achieves an 87% F1 score across seven public safety benchmarks, including Aegis and HarmBench. The model is designed to detect attacks such as narrative injection and encoding obfuscation, techniques commonly used to bypass AI safety guardrails.
Astroware Labs, a constitutional AI lab focused on guard models and agent security, operates Trishool as Bittensor Subnet 23. The team positions Halo as a production-grade safety system for AI products that need stronger defenses against jailbreaks and malicious prompt manipulation.
According to Trishool, the engagement begins as a pilot and is expected to evolve into a paid deployment protecting thousands of daily user interactions across Chutes products. The deployment will serve as a real-world test of Halo’s ability to operate at scale while defending against jailbreaks, exploits, and data security risks in user-facing AI applications.
Disclaimer: This article is for informational purposes only and does not constitute financial, investment, or trading advice. The information provided should not be interpreted as an endorsement of any digital asset, security, or investment strategy. Readers should conduct their own research and consult with a licensed financial professional before making any investment decisions. The publisher and its contributors are not responsible for any losses that may arise from reliance on the information presented.
