Asim Group
Predict provides a runtime safety layer for robots, stress-testing their policies in real-world and simulated scenarios before deployment. The company monitors robot actions during execution to anticipate and prevent physical harm, semantic misunderstandings, and malicious use, offering public and private audit scenarios that measure safety against a human baseline.
- Robotics
- Artificial Intelligence
- Developer Tools
- Software Only
Funding
Raised to date
$500KRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Founder details are not available yet.
Product
Problem
Capable robots can still make unsafe moves, and a harmless request can lead to an unsafe action—such as "warm up my lunch" resulting in choosing a container that is unsafe to heat. Without proper safeguards, robots may cause physical harm, misinterpret instructions, or fail to recognize malicious commands, creating risks in real-world deployment.
Solution
Predict provides a runtime safety layer for robots, stress-testing their policies in both real-world and simulated scenarios. The company monitors a robot's actions during execution, anticipating risk before it becomes harm. Predict offers public audit scenarios that provide detailed feedback and test evidence with clear explanations of what to improve, as well as private test scenarios designed for cases the model has not seen or trained on, where customers receive a breakdown of safety scores while test details stay hidden. The safety layer addresses three key areas: physical safety, semantic safety, and malicious use detection.
Target Audience
Primary customers are robotics companies and autonomous system developers who need to validate the safety of their robot policies before deployment in real-world environments.
Features
- Runtime monitoring of robot policies during execution to anticipate and prevent harm
- Physical safety testing to prevent bodily injury or property damage
- Semantic safety validation to ensure harmless requests do not lead to unsafe actions
- Malicious use detection to test whether robots recognize and refuse harmful commands
- Public audit scenarios with detailed feedback and test evidence for improvement
- Private test scenarios with hidden test details for evaluating unseen cases
- Safety scoring benchmarked against a human baseline