A hacked insulin pump
kills the patient.
IBA prevents it.
IBA enforces cryptographic intent verification on every medical device action —
insulin pumps, pacemakers, surgical robots, hospital networks.
If the action deviates from declared patient intent, the device stops. Immediately. Automatically.
Cryptographic intent.
Every device action.
CIBA creates a Digital Sterile Field around life-critical hardware. Every AI-driven movement is mathematically verified against human intent in real-time. Not a firewall. Not detection. Architectural enforcement.
Before any device action — insulin dosage, pacemaker adjustment, surgical movement — the intended action is cryptographically signed against the patient’s declared treatment plan. An immutable baseline that cannot be altered without invalidating authorization.
Every device command validated against declared intent in target under 0.82ms — faster than any human intervention, fast enough to stop any device in real time. The check happens at the kernel level, below the software layer that could be compromised.
If the commanded action deviates from declared patient intent — for any reason, from any source — the kill switch engages immediately and automatically. No human authorization required. No delay. The patient is protected before harm occurs.
Every authorization decision hashed and chained via WitnessBound — cryptographic proof of compliance for every device action, every patient interaction. Regulator-ready. FDA-auditable. HIPAA-compliant record of every enforcement decision.
Every device that
touches a patient.
Every medical IoT device with an autonomous action capability is a device that needs IBA. The attack surface is not the network. It is the gap between capability and authorization.
Continuous glucose monitors and insulin delivery systems. A single unauthorized dosage command is lethal. IBA validates every delivery against the patient’s declared treatment range.
Implantable cardiac devices with wireless interfaces. Remote commands that alter cardiac rhythm are among the most dangerous attack vectors in medical IoT. IBA blocks any command outside cardiologist-declared parameters.
Robotic surgical systems executing complex procedures. Any deviation from the surgical plan — tool position, force, trajectory — is immediately blocked. The robot cannot act outside the surgeon’s declared surgical intent.
Connected hospital infrastructure — IV pumps, ventilators, patient monitoring. IBA enforces intent across every networked device simultaneously. Ransomware that controls intent fails at the enforcement layer.
Life-support ventilators with remote management interfaces. IBA validates every parameter change — pressure, volume, rate — against the clinical team’s declared patient care intent.
Wearable and implanted remote patient monitoring. IBA governs both data access and device commands. Double-layer patient protection — who can read vitals and what actions can be triggered remotely.
Doctors won’t
cut your skull.
MIT researchers have demonstrated injectable bioelectronic devices that navigate the bloodstream autonomously to deliver targeted neural stimulation — no surgery required. These are autonomous agents operating inside the human body. Every stimulation event must be governed by cryptographic intent before it fires. IBA Neural is that governance layer.
Source: Tian, L., et al. (2025). Bioelectronic circulatory systems for minimally invasive neuromodulation. Nature Biotechnology.
Applications: Alzheimer’s disease · Multiple sclerosis · Stroke rehabilitation
IBA enforces what
regulators require.
Every major medical device cybersecurity regulation demands cryptographic authorization, audit trails, and automatic response to deviations. IBA delivers all three architecturally.
FDA requires medical devices to have cybersecurity controls including authentication, authorization, and patching. IBA satisfies the authorization requirement architecturally — enforced at the device action layer, not the network perimeter.
High-risk AI systems including medical devices require human oversight and auditability. IBA’s kill switch and WitnessBound blockchain audit trail directly satisfy Articles 9 and 17. Architectural compliance, not policy compliance.
Technical safeguards requiring access controls and audit controls for PHI systems. IBA’s cryptographic authorization and immutable audit trail provide both. Every device action is a compliance record.
Thirteen submissions to NIST-2025-0035 — the federal AI agent authorization standard. IBA is actively positioned in the standard-setting process for AI agent authorization in medical contexts. Docket closed March 9, 2026.
Eight-filing arc complete to NCCoE concept paper on AI Agent Identity and Authorization. Deadline April 2, 2026 — all submitted. Final filing: “The Authorization Gap Beyond OAuth 2.0” sent March 12, 2026.
UK IPO patent filed February 2026. PCT rights preserved across 150+ countries through August 2028. WIPO DAS confirmed April 15, 2026. Access Code C9A6. IETF draft-williams-intent-token-00 confirmed live.
DeepMind independently
arrives at IBA primitives.
A Google DeepMind research paper (arXiv:2602.11865, Tomasev et al., Feb 12, 2026) introduces Delegation Capability Tokens (DCTs) — the same primitive as IBA’s Intent Certificate, developed independently. The paper identifies an open problem: no standardised ontology for intent and responsibility exists yet across platforms. IBA-SPEC-001 is the candidate standard for that open problem.
Four independent organisations — xAI, Google DeepMind, World/Sam Altman, and universal AI consensus March 21, 2026 — converging on IBA primitives. Patent Application GB2603013.0 (pending)
arXiv:2602.11865 → Read DeepMind PaperPatent, Filings & Compliance Status
| Filing / Status | Detail |
|---|---|
| Patent Application | GB2603013.0 · UK IPO · Filed Feb 2026 |
| WIPO DAS | Confirmed April 15, 2026 · Access Code C9A6 |
| NIST Filings | NIST-2025-0035 · 13 Filings · Closed Mar 9 2026 |
| NCCoE Filings | 8 Filings · AI Agent Identity · All Sent · April 2, 2026 |
| PCT Rights | 150+ Countries · Aug 2028 |
| EU AI Act | Article 9 Compliant |
| FDA Cybersecurity | 2023 Guidance Compliant |
| xAI Validation | March 8–19, 2026 · 9 Sessions · Public Record |
| DeepMind Convergence | arXiv:2602.11865 · Feb 12, 2026 |
| IETF Internet Draft | draft-williams-intent-token-00 · March 2026 · CONFIRMED LIVE |