Home โ€บ Blog โ€บ The Maestro ยท Part 14
The Maestro ยท Part 14 ยท Phase 3 ยท ๐Ÿ›ก๏ธ Safety lead ยท ๐Ÿ’ผ founder hat

Approvals and Compliance

Maestro-1 works (Part 12) and is defended (Part 13). Before anyone outside the company uses it, Gopala has to answer to three audiences: regulators, customers' security and legal teams, and his own release gate. This part maps what the EU AI Act asks of a general-purpose model provider and where Maestro-1 sits (with a script that does the arithmetic), the voluntary frameworks customers ask about, the documents every release needs, customers' security reviews, and the release decision itself, which in the story first comes back no.

FL
FrontierAGI Team
๐Ÿ’ผ Founder๐Ÿ›ก๏ธ Safety leadโš–๏ธ Compliance๐Ÿ’ป Coder
โš ๏ธ How to read this (as of 30 Oct 2026) EU AI Act dates, the 10^25 FLOP presumption and provider duties are from the European Commission's published material; NIST AI RMF and ISO/IEC 42001 are cited from their publishers; model cards from Mitchell et al. (2019). ai_act_check.py and release_check.py were run (real output below); the release manifest is the scenario. model_card.md is an outline, not executed. This is not legal advice; rules change and have exceptions, so a real company needs counsel. Disclosure: Claude, made by Anthropic, helps write this site.

1. Mission log: weeks 50โ€“60

Weeks 50โ€“54 ยท Mapping
  • Maps obligations with counsel; writes ai_act_check.py; starts the documents.
Weeks 55โ€“58 ยท Documents and reviews
  • Model card, data summary, copyright policy, security questionnaires from three partners.
Weeks 59โ€“60 ยท Release reviews
  • First release review: no-go (section 10). Second, a week later: go, for design partners first.

2. Why approvals are a product feature

Three audiences, three kinds of evidence

๐Ÿ›๏ธ RegulatorsTechnical documentation, copyright policy, training-content summary
๐Ÿข CustomersSecurity questionnaires, contracts, model card, audit reports
๐Ÿšฆ The release gateEvaluation, safety, gate log, rollback, sign-offs

Much of the same evidence serves all three.

3. The EU AI Act and general-purpose models

4. Where Maestro-1 sits

ai_act_check.py ยท Python 3, no dependencies
"""ai_act_check.py โ€” where does Maestro-1 sit under the EU AI Act's
general-purpose AI (GPAI) rules? A planning aid, NOT legal advice.

Facts used: GPAI obligations apply from 2 August 2025; a model is
presumed to have systemic risk above 1e25 FLOPs of cumulative training
compute (rebuttable; the Commission can also designate models). Training
compute uses 6 x active params x tokens (Part 3); post-training adds a
little. Scale-up scenarios are hypothetical.
"""
THRESHOLD = 1e25
RUNS = {  # name: (active params, tokens)
    "Maestro-1 pre-training": (11.7e9, 10e12),
    "post-training + loop cycles (our estimate)": (11.7e9, 0.3e12),
}
BASE = ["technical documentation for the AI Office and downstream providers",
        "information and documentation for downstream providers integrating the model",
        "a policy to comply with EU copyright law, including opt-outs",
        "a public summary of the content used for training (Commission template)"]
SYSTEMIC = ["model evaluations including adversarial testing", "assess and mitigate systemic risks",
            "track and report serious incidents", "adequate cybersecurity protection"]

def main():
    total = 0
    for name, (n, d) in RUNS.items():
        f = 6 * n * d; total += f
        print(f"{name:44s} {f:.2e} FLOPs")
    print(f"{'cumulative':44s} {total:.2e} FLOPs = {total/THRESHOLD:.1%} of the 1e25 presumption threshold")
    print("obligations as a GPAI provider placing the model on the EU market:")
    for o in BASE: print("  -", o)
    for mult in (5, 10, 15):
        f = total * mult
        print(f"if a future run used {mult:2d}x the compute: {f:.1e} FLOPs -> "
              f"{'presumed systemic risk: add ' + str(len(SYSTEMIC)) + ' more duties and notify the Commission' if f > THRESHOLD else 'still below the threshold'}")

if __name__ == "__main__":
    main()
$ python3 ai_act_check.py # planning aid, not legal advice Maestro-1 pre-training 7.02e+23 FLOPs post-training + loop cycles (our estimate) 2.11e+22 FLOPs cumulative 7.23e+23 FLOPs = 7.2% of the 1e25 presumption threshold obligations as a GPAI provider placing the model on the EU market: - technical documentation for the AI Office and downstream providers - information and documentation for downstream providers integrating the model - a policy to comply with EU copyright law, including opt-outs - a public summary of the content used for training (Commission template) if a future run used 5x the compute: 3.6e+24 FLOPs -> still below the threshold if a future run used 10x the compute: 7.2e+24 FLOPs -> still below the threshold if a future run used 15x the compute: 1.1e+25 FLOPs -> presumed systemic risk: add 4 more duties and notify the Commission

Maestro-1's training compute vs the systemic-risk presumption

Maestro-1 (pre-training + post-training)7.2e23 FLOPs (7%)
10ร— Maestro-17.2e24 (72%)
15ร— Maestro-11.1e25 (above)
Threshold1e25 FLOPs

Real output of ai_act_check.py (6ยทNยทD estimate, 11.7B active parameters).

5. Voluntary frameworks customers ask about

FrameworkWho publishes itWhat it isHow Gopala uses it
AI Risk Management Framework 1.0NIST (US), January 2023Govern, map, measure, manage AI risksStructure for the risk register and questionnaires
ISO/IEC 42001ISO/IEC, December 2023Certifiable AI management system standardPlanned for year two; certification is costly for a start-up
SOC 2 Type IIAICPAAudit of security controls over timePartly inherited from the GPU provider (Part 4); own audit in year two
OWASP LLM Top 10OWASPApplication security risks for LLM appsChecklist behind Part 13

6. The documents every release needs

model_card.md ยท model card outline ยท not executed for this article
# Maestro-1 โ€” model card (outline)
## Model
- Developer: Maestro (start-up) ยท Version: 1.0 ยท Date: month 13
- Architecture: mixture-of-experts, 65.2B total / 11.7B active parameters
- Training compute: ~7.2e23 FLOPs (6ยทNยทD estimate) ยท Context: 64K tokens
## Intended use
- Software-engineering assistance inside customers' own repositories, through agents with a tool guard
- Not intended for: unsupervised production changes, medical/legal advice, decisions about people
## Training data
- ~10T tokens: filtered web, permissively licensed code, maths, books, multilingual, verified skills
- Public summary: see data-summary.pdf (EU template) ยท opt-outs honoured
## Evaluation (harness v3, contamination-checked, 95% CIs)
- Partner held-out tasks: pass@1 ... [CI] vs Maestro-0 ... [CI]
- Public benchmarks: ... (with contamination notes)
## Safety
- Red-team attack success by category with Wilson intervals; known weaknesses listed
## Limitations
- Below frontier models on general knowledge; English-first; can still write insecure code
## Contact
- security@ and incidents@ addresses, response times

7. Customer security reviews

8. The release review

release_check.py ยท Python 3, no dependencies; scenario manifest
"""release_check.py โ€” the release gate for a model version.

Every item must exist and be signed before Maestro-1 ships to anyone.
The manifest below is the SCENARIO state at the first review: two items
are missing, so the answer is no.
"""
REQUIRED = {
    "model_card": "model card: intended use, limits, evaluation results with CIs",
    "eval_report": "evaluation report from the harness (Part 12), contamination checked",
    "safety_report": "safety evaluation and red-team summary (Part 13)",
    "data_summary": "public training-content summary (EU template) and data manifest",
    "copyright_policy": "copyright and opt-out policy",
    "gate_log": "loop gate decisions since the last release (Part 11)",
    "incident_process": "incident contact, severity levels, response times",
    "partner_signoff": "design partners reviewed on their own data",
    "rollback_plan": "previous version kept serving; rollback tested",
}
MANIFEST = {
    "model_card": {"done": True, "signed": "Gopala"},
    "eval_report": {"done": True, "signed": "Gopala"},
    "safety_report": {"done": True, "signed": "external red team + Gopala"},
    "data_summary": {"done": False, "signed": ""},
    "copyright_policy": {"done": True, "signed": "counsel"},
    "gate_log": {"done": True, "signed": "Gopala"},
    "incident_process": {"done": True, "signed": "Gopala"},
    "partner_signoff": {"done": True, "signed": ""},
    "rollback_plan": {"done": True, "signed": "Gopala"},
}

def main():
    missing = []
    for k, desc in REQUIRED.items():
        m = MANIFEST.get(k, {})
        ok = m.get("done") and m.get("signed")
        print(f"{'OK  ' if ok else 'MISS'} {desc}" + ("" if ok else f"  ({'not done' if not m.get('done') else 'not signed'})"))
        if not ok: missing.append(k)
    print("RELEASE: " + ("GO" if not missing else f"NO-GO, {len(missing)} item(s) open: {', '.join(missing)}"))

if __name__ == "__main__":
    main()
$ python3 release_check.py # scenario state at the first review OK model card: intended use, limits, evaluation results with CIs OK evaluation report from the harness (Part 12), contamination checked OK safety evaluation and red-team summary (Part 13) MISS public training-content summary (EU template) and data manifest (not done) OK copyright and opt-out policy OK loop gate decisions since the last release (Part 11) OK incident contact, severity levels, response times MISS design partners reviewed on their own data (not signed) OK previous version kept serving; rollback tested RELEASE: NO-GO, 2 item(s) open: data_summary, partner_signoff

9. Try the release gate

Release checklist (same items as release_check.py)

10. A staged release

Who gets Maestro-1, and when

๐ŸคDesign partnersMonth 13.5; their own repositories, close support
๐ŸงชPilot customersMonth 14.5; limited seats, weekly reviews
๐ŸŒGeneral availabilityMonth 16+; only after pilot evidence

Each stage has its own gate; any severity-1 incident pauses the next stage.

11. Personal data in customers' code

12. One evidence library

From records to documents

๐Ÿ—‚๏ธRecordsManifests, gate log, eval runs, access logs
๐ŸงพControlsEach control written once with its evidence
๐Ÿ“„DocumentsModel card, data summary, policies
๐Ÿ“จAnswersRegulators, customers, auditors

Write the record when the work happens; documents become assembly, not archaeology.

13. The compliance calendar

WhenWhat
Every releaseRelease check, model card update, data summary update if training data changed
Every loop cycle (2 weeks)Gate log entry with sign-off (Part 11)
QuarterlySafety drill, red-team refresh, review of new regulatory guidance
YearlyExternal security audit; review of the obligation map with counsel
Before any larger training runRe-run ai_act_check.py; plan for systemic-risk duties if near the threshold

14. Clauses that carry the promises

ClauseWhat it saysBacked by
Data useRaw customer code is never used for training; abstracted skills only with consentPart 8 pipeline and manifest
RemovalSkills derived from a customer's tasks are removed on request from future trainingManifest and skill-library ids
ResidencyProcessing in named regionsSandbox and provider configuration
IncidentsNotification within stated hoursPart 13 incident process
Sub-processorsGPU and cloud providers listed; changes notifiedPart 4 contracts
Model changesNotice before model updates; rollback availablePart 11 gate, Part 15 rollout

15. Beyond the EU

16. Who owns compliance as the company grows

17. Seven approval mistakes

18. Delegation box

๐Ÿ™‹ Gopala himself
  • The release decision
  • Positions on data use in contracts
  • The obligation map with counsel
๐Ÿค– AI agents
  • Drafting documents from the manifest and reports
  • First drafts of questionnaire answers
  • Tracking regulatory updates
๐Ÿค Bought, rented or borrowed
  • Outside counsel (EU and contracts)
  • GPU provider's audit reports
  • A compliance tool for questionnaires

19. Budget and risk tracker

ItemStatus after Part 14Note
Spend to date~$7.4M (scenario)Legal, documentation, reviews
StatusReleased to design partners (scenario)Second review: go
TimeWeek 60 (month ~14)Phase 3 complete
Regulatory riskManagedBelow 1e25; baseline GPAI duties met
Customer-trust riskWatchQuestionnaires answered; audits in year two
Future-scale riskPlanMaestro-2 at ~15ร— compute would cross the threshold

20. The series

The Maestro: 18 parts in five phases

Verdict
  • Phase 3 ends with evidence: a harness that can say no, six safety layers, a document set built from records rather than memory, and a release gate that did say no once before saying yes.

21. Reading the numbers

Primary / standard
EU AI Act GPAI obligations from 2 Aug 2025; legacy models until 2 Aug 2027; enforcement from 2 Aug 2026; 10^25 FLOP presumption; Code of Practice finalised 10 Jul 2025 (European Commission). NIST AI RMF 1.0 (Jan 2023). ISO/IEC 42001 (Dec 2023). Model cards (Mitchell et al., 2019).
Secondary โš ๏ธ
Some secondary summaries of the Code of Practice were used to check dates (โš ๏ธ); the Commission pages are the reference.
Our assumptions
The release manifest, review outcomes, timelines, costs and scale-up scenarios.