Financial institutions are adopting artificial intelligence at speed, and the direction is not in question. Models and agents are moving into credit, surveillance, onboarding, claims, and customer contact. Within a few years, every institution of consequence will run AI in some form.
An institution must be able to answer for what its AI is doing.
"If you don't understand the system you're using, you don't control it. If nobody understands the system, the system is in control." Philip Monk, Precepts
The burden of that answer lands on the compliance and risk function. Chief compliance officers are being asked to extend risk management to systems that did not exist when their control frameworks were written, and to do it on supervisory timelines. The people in these seats need new controls, new evidence, and new skills, quickly.
A new role is forming in response. Some institutions fold AI governance into the chief compliance officer's mandate. Others are creating an AI governance officer who reports to the chief compliance officer. The role is new, the title varies, and no shared definition of the job exists: what it manages, what it must be able to evidence, what good looks like.
The officers taking on this accountability, which in several jurisdictions carries personal regulatory accountability when the mandate sits with a senior manager, have no peer reference. There is no pooled data on how AI risk programs are staffed, which controls withstand examination, or how often things go wrong. That data exists inside every institution and is pooled nowhere. Each officer solves the same problems alone, and the loudest available guidance comes from vendors with something to sell.
The AI Risk & Compliance Network is established to close that gap: a confidential benchmark of how these programs actually operate, a peer body for the people who run them, open reference materials that define the role and its standards, and closed-door programming that builds the profession's capability.
This charter is the Network's constitution. It is public, versioned, and binding on Future Native, Inc. ("Future Native"), which operates the Network during its formation. Where this charter and any commercial interest of Future Native conflict, this charter prevails.
1.1 To establish the officers accountable for AI at financial institutions as a profession with a peer body: to convene them, under the Chatham House Rule, so that the people carrying this new accountability can compare practice candidly and develop the standards of their field together.
1.2 To operate a confidential, contribution-gated benchmark of how AI risk and compliance programs at financial institutions actually operate.
1.3 To publish the reference materials of the profession: the Network's reporting taxonomy and standards, the reference definition of the AI governance role, and an annual report of headline aggregate findings, all free and open. Deeper research derived from the pool is published under §3.5.
1.4 To build the profession's capability through closed-door member programming: peer briefings, practitioner workshops, and roundtables held under the Chatham House Rule.
1.5 The Network is one global body with country chapters, each chapter mapped to its local regulatory context and led by a locally trusted convener. One benchmark and one standard, convened locally.
1.6 The Network anchors on the permanent operating needs of its members (peer comparison, incident intelligence, professional definition, capability building) and not on any single law, regulation, or supervisory instrument.
2.1 By function, not title. Membership is open to the person who owns or co-owns the AI accountability function at a financial institution, whatever their title, subject to admission under §2.2. Membership is institutional, exercised through a named officer. There is no individual or vendor membership.
2.2 Admission requires contribution. Every admitted institution must: (a) submit an institution profile at admission (the benchmark's baseline module); (b) demonstrate the ability to contribute data in line with the Network's reporting standards; and (c) formally attest to the quality and completeness of the data it will supply.
2.3 Dues. Members pay annual dues set at cost-recovery levels (§6.5), scaled to institutional size by group annual revenue band, self-certified at admission as part of the member's attestation. Founding-cohort rates: Emerging (<US$50M): US$1,500 · Mid-size (US$50M–1B): US$3,000 · Major (>US$1B): US$6,000. A misstatement of band is remedied by reclassification. Dues reflect ability to pay and nothing else: governance is one member, one vote, and no tier purchases influence over the methodology or the standard. Dues are never based on data reported into the benchmark. Contribution of data under §4 is a condition of membership and is not a substitute for dues, nor dues for contribution. Founding rates are deliberately low. They price a Network that is still building its pool, and they are locked for founding members through the pilot period. Dues are revisited with version 2 of this charter, once the founding baseline exists, and are expected to rise toward peer-association levels as the accumulated benchmark becomes what later joiners buy into. A one-time joining contribution for post-founding institutions will be introduced at the same time. Any interim surplus is held in a disclosed reserve for the Network's audit and eventual independence (§8), per §6.5.
2.4 Confidentiality among members. Proceedings are held under the Chatham House Rule. Member identities, benchmark contents, and closed-door discussions are confidential under a mutual NDA. No member ever sees another member's raw or attributable data, only the aggregate and its own position within it.
2.5 The founding cohort. The Network's first year is explicitly a founding pilot cohort. Its outputs are the founding-member baseline, described as such and never as "the industry."
3.1 The Network's reporting taxonomy and standards are published openly under a Creative Commons (CC-BY 4.0) license. Anyone, member or not, may adopt them without joining or paying.
3.2 The reference definition of the AI governance role (what the officer accountable for AI should manage and be able to evidence) is published free, and is grounded in what the benchmark shows the field actually does.
3.3 An annual report of headline aggregate findings, non-attributable, is published free.
3.4 Adopting or implementing the standard never requires membership, payment, or the products of any vendor, including Future Native.
3.5 Research publications. The Network may publish deeper research derived from the pool (periodic analyses, thematic studies, survey findings) as a paid subscription publication (working name: The Baseline) available to non-members; members receive it as part of membership. All such publications are aggregate and non-attributable, subject to §4.4 and §5, and their revenue is Network revenue under §6.5. The materials in §§3.1 through 3.3 are never paywalled.
4.1 The confidential benchmark, showing where an institution stands against its anonymized peers, is derived from member contributions and available only to contributing members. Contribution buys a view of your own position. It never buys influence over the standard.
4.2 Ratio-based reporting. To make a small pool comparable, the Network collects scaling denominators from day one: the count of AI models and use-cases in production, and the volume of AI-driven decisions or agent actions. It reports in ratios, not raw counts.
4.3 Anonymized rank positions. Members receive their own rank position against the pool ("you are position 4 of 18; pool median is X"), never another member's identity or data.
4.4 Small-cell honesty. No statistic is reported from fewer than five contributing members, or seven for the most sensitive fields. What cannot be reported at the current pool size is stated, not approximated.
4.5 Cadence. Contributions run as an annual full wave, with a quarterly incident-and-findings pulse between waves, quality-checked against pool patterns before release.
4.6 Anchor stability. The benchmark's anchor questions and band boundaries are frozen for a minimum of three cycles. Changes are additive and versioned, governed by §7.3.
4.7 Benchmark IP. The aggregate benchmark and the statistics derived from it belong to the Network, licensed to members for their internal use. Each member retains ownership of its raw contributions and grants the Network a limited, purpose-bound aggregation license under §5.
5.1 Members own their data. Each contributing institution remains the owner and data controller of its raw contributions. The Network holds only the enumerated usage rights in this section, and Future Native acts solely as processor on documented instructions.
5.2 Purpose limitation. Contributed data is used only to (a) operate the benchmark and (b) inform the public, non-attributable reference materials in §3. It is not used for any other purpose.
5.4 De-identification at ingestion. Identifiers are stripped at intake under a documented procedure. The aggregate store holds no member-identifiable data. Future Native's commercial function has access to neither the raw nor the identifiable layer.
5.5 Audit. The confidentiality system (intake, storage, access, suppression) is subject to an annual independent review, and members hold audit rights to verify the treatment of their own contributions.
5.6 Exit. On withdrawal, a member's raw contributions are deleted. Already-published, non-attributable aggregate statistics persist.
5.7 Regulator-derived data. The Network does not solicit examination or supervisory-confidential information unless and until counsel confirms, per jurisdiction, that members may lawfully contribute it. The benchmark launches on self-assessed control, incident, and governance data.
6.1 No sales funnel. The Network's roster, proceedings, and benchmark participation are never used by Future Native as sales leads. Future Native's commercial staff may not initiate commercial contact with any person in their capacity as a member.
6.2 Inbound only, logged. A member that wishes to engage Future Native commercially may do so only through a documented, member-initiated request. Such engagements are logged and disclosed to the Network's governance in aggregate ("N members engaged commercially this period").
6.3 Sold on the standard, never on the score. Any Future Native product is sold on the open standard of §3. Future Native never uses a member's benchmark position in any commercial conversation.
6.4 Honest scale. During formation, the Network is two people and a charter. The separations in this section are commitments of conduct: written, logged, and auditable. They will become separations of personnel as the Network grows. This charter states that plainly rather than claiming walls that do not yet exist.
6.5 Cost recovery, separate books. Dues and any Network fees, including fees for member programming under §1.4, fund the Network's operations (convening, the benchmark, programming, publication) on a separate, auditable ledger. The Network is not a profit center for Future Native.
6.6 No vendor sponsorship in the founding phase. Any future sponsor tier requires the governance of §7 and may never include access to members, member data, or the benchmark.
7.1 Conveners. The Network was founded by Daniel Hwang and is convened with Angelina Kwan: title per canonical bio; sign-off confirmed, wording pending. Future Native, which Daniel co-founded, supports the Network's operations under this charter, an affiliation disclosed here and wherever the Network is described.
7.2 Member-governed methodology. The benchmark's methodology (what is measured, how it is scored, and the anonymization thresholds) will be governed by the members through a Steering Committee, majority non-Future-Native with an independent chair. The Committee's composition, seating, and procedures are defined in version 2 of this charter, to be adopted with the founding cohort.
7.3 Schema evolution. Between charter versions, changes to the taxonomy and reporting standards are proposed by a small technical committee and adopted after member comment, additive and versioned, per §4.6.
7.4 Chapters. Each country chapter operates under this charter with a named local convener. Chapters localize language and convening. The substance (taxonomy, bands, anchors) stays globally invariant.
8.1 Future Native publicly commits to move the standard and the benchmark into an independent, non-profit or at-cost entity with its own governance when any of the following occurs:
8.2 On spin-out, the commitments of §§3–6 transfer to the successor entity intact.
9.1 This charter is public at airc.network/charter, versioned, and dated. Amendments are logged with rationale. The commitments in §5.3, §6.1, and §8 may be strengthened by amendment but not weakened.