BYSS Labs
Three labs and one pilot product. Each does something its category does not, and proves it with measurements you can rerun yourself.
Every claim comes with what you need to check it.
AEGIS decides whether an AI agent may act, and proves it with a receipt that an independent verifier rereads. MAIVAR reads an entire Windows workstation in twenty seconds without writing a byte, down to what an AI assistant can launch there. Caribbean BioAI shows, across 3,731 species, that a plant's described chemistry follows human attention, not rarity. Passeport Machine keeps a machine's regulatory file as a living register, and refuses to print a false one. In all four, what is claimed can be redone, and what was ruled out stays on the record.
A control layer that refuses to let an AI agent act until its identity, its mission, the evidence, the authorization and a safe rollback can all be verified.
A complete snapshot of a Windows workstation in twenty seconds, without admin rights and without writing a single byte, including what an AI assistant can launch on your behalf. Sealed, signed, and anchored in the AEGIS transparency log.
A reproducible study of the flora of Martinique shows that a plant's described chemistry follows human attention, not rarity. It keeps every hypothesis it ruled out, with the reason why.
The regulatory twin of a machine under Regulation (EU) 2023/1230: a digital instruction manual behind a QR code, a versioned technical file, an evidence register, and a watch on the applicable texts.
AI assistants now launch programs on employees' machines, with their privileges. What shadow AI is, and how to find it on your endpoints: the three forms to look for, why classic security tools miss the newest one, and what EU law already requires.
