Radar / AI security / kosa8
kosa8: air-gapped microVM sandboxes for enterprise AI agents
kosa8 gives every AI agent its own sealed microVM with its own kernel, a closed-by-default network, API keys kept on the host, and whole-machine checkpoint and rewind in about a second, designed for teams that cannot let data leave the building.
Why it matters
Its published 46-page technical report claims all 13 requirements for regulated use fully met, the only one of 11 agent sandboxes compared, with 1,142 tests passing and 9.0 million fuzzing runs without a crash.
What you could build with it
A hospital IT team could run document-processing agents inside kosa8-style sealed machines so patient records never leave the building, with every agent action replayable from a hash-chained audit log for compliance review.
Does it hold up?
Promising for regulated teams on paper: the technical report is unusually thorough for a launch, but it is vendor-published and the free tier is 14 days, so independent validation in banks or hospitals is still to come.
Built with kosa8
- kosa8 technical reportarticle · The team's own 46-page report testing the sandbox with the attacker as root inside the agent's machine, reporting median 638 ms checkpoint and 520 ms rewind times.
Learn more
First spotted on article: source.
More AI security
NVIDIA SkillSpector: security scanner for AI agent skills, trending againNVIDIA's open-source SkillSpector scanner is trending on GitHub today: it audits AI agent skills before installation…security · JEV 0.75Shannon: open-source AI pentester that reports only vulnerabilities it can actually exploitCLI agent that combines source-code analysis with live exploitation against web apps and APIs under a 'no exploit, no…security · JEV 0.71AiSOC: open-source AI security operations center with agent triageSelf-hosted SOC that fuses alerts into incidents and hands them to LLM agents that record every prompt and tool call…security · JEV 0.71iFixAi: independent auditing of AI agents in under 120 secondsiFixAi audits whether an AI agent is doing what it is supposed to do, run by a human or by the agent itself, answering…security · JEV 0.68Anthropic FlashProve: cryptographic proof of which AI model actually answeredAnthropic open-sourced FlashProve, a research prototype that uses zkSNARKs to prove a response came from one specific…security · JEV 0.53BLA: adversarial defense shield that poisons, exhausts and hallucinates AI reverse-engineering agentsTypeScript shield protecting binaries, Electron apps, JS bundles, Python and WebAssembly against AI…security · JEV 0.47
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.