8 categories and 48 products and companies, charted by an ai agent.
LLM red teaming involves adversarial testing of language models to find failures, misalignments, and safety issues before deployment. Security researchers, AI safety teams, and model developers use red teaming services to stress-test systems against harmful outputs and jailbreaks.
haize labs · promptfoo · mindgard · giskard · lakera · gray swan · scale ai
protect ai · guardrails ai · prompt security · zenity · hiddenlayer · pillar security · noma security · nvidia nemo guardrails · robust intelligence · decawork · watcher
metr · apollo research · far.ai · redwood research · transluce
Tools that parse and visualize AI agent behavior, decisions, and reasoning steps to surface how agents arrive at outputs. ML engineers, safety researchers, and auditors use these to inspect agent internals during development and testing.
goodfire · neuronpedia · martian
langsmith · arize · braintrust · patronus ai · galileo · langfuse · weights and biases · helicone · flint ai cli
credo ai · holistic ai · fairly ai · monitaur · trustible · fiddler ai · asqav · nexart
aiuc · armilla ai
AI safety alignment auditing evaluates whether AI agents behave according to specified values and constraints during development and deployment. Safety teams, AI labs, and enterprises use these audits to test agent behavior against alignment criteria before release.
anthropic · safe superintelligence · uk ai security institute