§ Standards
We publish the standards we're held to.
A category is only real when its rules are public. Summit's specifications are open under CC BY 4.0, versioned like software, and written to be adopted — by customers, by competitors, by procurement offices, by anyone who believes machine-generated decisions should come with proof.
§ 01
The four anchors
v1.0 · CC BY 4.0
Decision Receipt Specification→
The receipt itself: structure, deterministic evidence-ID format, validation rules, and trust model. If you implement nothing else of ours, implement this.
v1.0 · CAL-1 → CAL-7
Cognitive Security Framework→
Seven functions from IDENTIFY to ADMIT, seven assurance levels, a five-level maturity model — and §7: procurement language ready to paste into an RFP.
v0.1 · benchmark
AGD-Bench Methodology→
Verified Governed Delivery Score (VGDS): the first benchmark that measures whether agentic work can be proven, not just whether it can be done.
v1.0 · incident catalog
Decision Failure Atlas→
Documented, public AI decision failures — sanctioned attorneys, withdrawn chatbots, exfiltrated secrets — each mapped to the control that would have prevented it.
§ 02
Participation, not just publication
Standards work is credibility work. Summit files public comments, submits to consortia, and certifies against its own specs.
Adopt the standard before you buy anything.
Every spec is free to implement. The procurement language in CSF §7 works whether or not Summit is the vendor — that's the point.