KYA Standard v1.5 — what changed since the v1 field note
· by Risto Anton Paarni
Version clarification (read this first)
- Public standard name: KYA Standard v1.5 (the document title in
Legal/KYA_STANDARD_v1.md). - Internal document metadata: Version 1.7 (the latest edit in the version history, 2026-04-05).
- Split documents: Know Your Agent (KYA-S v1.5, CISO focus) and Know Your Architecture (KYA-O v1.7, CFO focus). Same three letters; zero content overlap.
- Original field note: KYA Standard v1 (9 March 2026) remains online as the historical record.
Four changes turned the March 2026 field note into the standard we operate today.
Change 1 · New in v1.4
Pillar 4 — Autoresearch & Agent Self-Optimization Governance
KYA v1 had three pillars: Identity Attribution, Capability Gating, Forensic Observability. v1.4 added a fourth. Autoresearch lets agent swarms iterate on their own models, policies and proxy metrics under hardware-attested safety bounds. Every autoresearch run is logged, scored and attributed. The Agent Trust Score formula now carries an explicit autoresearch delta: 1.0 − V×0.25 − L − D×0.05 + A×0.15.
Change 2 · New in v1.5
Pillar 5 — KYA Vision: AI Orbital Target Labeling
Before governing an agent, know what it is governing. KYA Vision adds eyes. ICEYE SAR satellites supply the imagery (60+ satellites, sub-daily revisit, 16 cm Gen4 resolution, all-weather). The Argus agent classifies targets across 20 industries under KYA governance — every label gets a Firehorse trace ID, low-confidence labels route to human review, wedge industries always route to human review. Finnish constellation, Finnish governance; same regulator.
Change 3 · New in v1.4
The Agentic Efficiency Law and the Stacking Effect
Two named laws replace hand-wavy productivity arguments. The Agentic Efficiency Law: for every 1 % improvement in autoresearch iteration speed, measurable industrial OPEX falls 0.7–2.2 % depending on sector. The Stacking Effect: governed autoresearch gains compound across agents in a swarm, so aggregate efficiency outruns single-agent benchmarks. Benchmarked 11–30 % efficiency gains from real autoresearch runs; projected 8–22× ROI within 18 months in heavy industry.
Change 4 · v1.4 financial layer · v1.5 industry count
Financial Value Model across 20 industries
v1 named the architecture. v1.4 added a per-industry revenue model: verification volume economics, insurance and liability reduction, carbon-penalty avoidance ROI, and a Trust Score table with a financial multiplier column. v1.5 widened coverage from 16 to 20 industries (added Fintech, Pharma, Cybersecurity, Healthcare as wedge industries). Aggregate projection: €14–38 billion EU-wide impact by 2030.
Also since v1
- Subagent Registry & Capability Manifest v1.2 — path-scoped access control, fault attribution in Supabase, KYA Violation enforcement (v1.1, refined in v1.4).
- Orchestrator Layer — kill early, promote winners. Swarm management stopped being manual.
- Memory audit trail — KYA-S v1.5 §10 and KYA-O v1.7 §12 govern dual LWM + client memory with fact-extraction lineage (v1.6 metadata).
- Article 99(4) correction — Regulatory Penalty Avoidance now cites Art. 99(4) (€15 M / 3 % for high-risk deployer non-compliance), with Art. 99(3) reserved for Art. 5 prohibited-practice context (v1.7 metadata, 2026-04-05).
Where the canonical v1.5 standard lives
The source of truth is Legal/KYA_STANDARD_v1.md in the Lifetime Oy repository — document title KYA Standard v1.5, metadata Version 1.7. The split documents for specialist audiences are Legal/KYA_SECURITY_v1.2.md (KYA-S, CISO) and Legal/KYA_OPTIMIZATION_v1.4.md (KYA-O, CFO). A NIST submission set (KYA_NIST_STANDARD_v1.md, KYA_NIST_SECURITY_v1.md, KYA_NIST_OPTIMIZATION_v1.md) derives directly from these.
For the original architectural case — why software-level agent scaffolding is insufficient for EU-regulated industries, why hardware-enforced isolation matters, how Firecracker MicroVMs became the bedrock — read the v1 field note. It is still the right starting document. v1.5 extends it; it does not replace it.
Risto Anton Paarni
CEO, Lifetime Oy · Editor in Chief, Lifetime Scope Journal