AI-First Exchange (AIFX) is an open, governance-first container standard for AI-generated media.
It defines how AI-created works are packaged, declared, and verified — focusing on provenance, integrity, and portability, not ownership claims or legal adjudication.
See VISION.md for long-term direction and roadmap.
As AI-generated content becomes ubiquitous and AI answers replace traditional search clicks, creators and platforms need a clear way to answer:
- Who is declaring authorship?
- How was this work generated?
- Has this package been altered since it was created?
AIFX provides a defensible, machine-verifiable structure to answer those questions.
AIFX is a container format family that bundles:
- AI-generated media assets
- A structured
manifest.json - Declared authorship and creation context
- Cryptographic hashes for integrity verification
AIFX packages are designed to be:
- Portable across platforms
- Verifiable without central authority
- Honest about what they do and do not claim
- ❌ Not a copyright system
- ❌ Not proof of originality
- ❌ Not an identity verification authority
- ❌ Not a moderation or enforcement platform
AIFX records Self-Declared Authorship (SDA) and enforces structural and integrity rules only.
Trust scoring, identity verification, and enforcement belong to downstream platforms.
- SDA only — Self-Declared Authorship is the sole verification tier in v0
- Validators do not lie — packages deterministically pass or fail
- No silent assumptions — required fields are enforced explicitly
- Structure before UX — governance precedes convenience
- Provenance over platforms — packages must outlive services
- AIFM — AI-generated music
- AIFV — AI-generated video
- AIFI — AI-generated images
- AIFP — AI-generated projects
All formats share the same provenance and integrity model.
- aifx-desktop
Converter and validator desktop tooling for AIFX packages.
Additional repositories will appear as components stabilize.
Using AIFX CLI in CI/CD pipelines
The aifx CLI is designed for automation and build systems.
Validation supports machine-readable output and CI-safe exit codes.
Exit codes
0 — All packages PASS
2 — One or more packages FAIL validation
1 — CLI or input error
This allows validation to gate builds and deployments.
python -m aifx validate ./artifacts
Emit machine-readable JSON (for automation)
To print JSON to stdout (quiet mode implied for clean output):
python -m aifx validate ./artifacts --json
To explicitly suppress human-readable output:
python -m aifx validate ./artifacts --json --quiet
Write JSON report to file (recommended)
python -m aifx validate ./artifacts \
--json \
--json-path ./aifx-validation.json
This preserves human output in the terminal while writing structured results to file.
⸻
Example: fail build on validation error
python -m aifx validate ./artifacts --json --json-path report.json
code=$?
if [ $code -ne 0 ]; then
echo "AIFX validation failed (exit=$code)"
exit $code
fi
⸻
Example JSON structure
{
"tool": "aifx-cli",
"timestamp": "...",
"input_path": "...",
"totals": {
"count": 2,
"pass": 2,
"fail": 0
},
"results": [
{
"package": "...",
"valid": true,
"errors": [],
"warnings": [],
"checks": { ... }
}
]
}
⸻
AIFX is under active development.
Expect:
- refactors
- evolving specifications
- breaking changes
What is stable:
- governance philosophy
- verification boundaries
- integrity enforcement model
The AIFX Desktop tooling implements and validates against the canonical format specifications maintained in the AI-First-Exchange Standard repositories.
Authoritative specs:
-
AIFM (Music)
https://github.com/AI-First-Exchange/AIFM-Standard/tree/main/spec -
AIFV (Video)
https://github.com/AI-First-Exchange/AIFV-Standard/tree/main/spec -
AIFI (Image)
https://github.com/AI-First-Exchange/AIFI-Standard/tree/main/spec -
AIFP (Project)
https://github.com/AI-First-Exchange/AIFP-Standard/tree/main/spec
The Desktop validator mirrors the canonical validation check names defined in each format’s VALIDATION_CHECK_NAMES_V0.md.
Specifications define structure and integrity requirements.
Tooling enforces those requirements deterministically.
Specs are authoritative.
Tooling must not drift.
AI First Exchange™ (AIFX™) is an open standard stewarded by its original author and community contributors.
Use of the source code or specifications does not grant permission to imply official endorsement or certification beyond what is explicitly documented.
Declare only what you can prove today.
Design for what you can verify tomorrow.