FractalAI is the first AI-native Layer 1 blockchain built from scratch in Rust. It features post-quantum cryptography (CRYSTALS-Dilithium), Proof of Fractal Work consensus, WASM smart contracts, and 460+ RPC endpoints.

What is the FRAC token?

FRAC is the native utility token of the FractalAI blockchain with a fixed supply of 1 billion tokens. It is used for staking, governance, gas fees, and DeFi. FRAC is purely deflationary with a 61.8% fee burn rate based on the golden ratio.

How does Proof of Fractal Work differ from Proof of Work?

Proof of Fractal Work (PoFW) requires validators to perform fractal analysis instead of hash puzzles. Validators whose fractal dimensions approach the golden ratio (φ = 1.618) receive higher rewards, making it computationally meaningful rather than wasteful.

Is FractalAI quantum-resistant?

Yes. FractalAI uses CRYSTALS-Dilithium for digital signatures and Kyber for encryption, both NIST-approved post-quantum cryptographic algorithms that protect against quantum computing threats.

How can I buy FRAC tokens?

FRAC tokens are available through the presale at fractalai.net.co/presale. You can pay with USDC on Base, USDC on Stellar (lowest fees), or credit/debit card via Circle. The presale has 4 phases starting at $0.25 (Seed) up to $0.80 (Public), with listing target at $1.00.

What is the AURUM protocol?

AURUM is FractalAI's commodity tokenization protocol that enables gold, silver, platinum, and 5 other commodities to be tokenized on-chain with phi-bonding curves that grow at 1.618% per epoch.

← Learn

#eu ai act article 50#ai transparency obligation#ai content labeling

How to comply with EU AI Act Article 50 (AI transparency obligations)

Article 50 of the EU AI Act sets transparency obligations for AI: providers of generative systems must ensure their synthetic audio, image, video and text output is marked as artificially generated in a machine-readable, detectable format, and deployers must disclose deepfakes and AI-generated text published to inform the public. The obligations apply from August 2026, and unlike the high-risk rules they reach a very broad set of everyday generative uses.

What the obligation actually requires

Article 50 does not ask you to prove content is real — it asks you to disclose when content is AI-generated or manipulated. Providers must embed a machine-readable marker in generated output; deployers of deepfake and AI-text systems must label content presented to the public. The marker has to be robust and interoperable "as far as technically feasible," which points toward open standards rather than a proprietary watermark only your own tools can read.

Marking versus detection

The compliant path is to mark at generation time, not to detect after the fact. AI detectors that guess whether text or an image is synthetic are unreliable and are not what Article 50 asks for. A provenance marker — a signed manifest bound to the asset declaring it was AI-generated, by which tool, when — is machine-readable, verifiable and durable. Anyone can check the marker; you are not depending on a probabilistic classifier that can be fooled or that flags human work as machine work.

Honest scope

Provenance marking satisfies the transparency and disclosure obligation; it is not a full AI Act compliance program, which for high-risk systems also spans risk management, data governance and human oversight. A marker proves an asset carries a signed declaration and has not been altered since — it does not detect unmarked content someone else generated, and it is not an AI detector. Sign markers with a durable scheme so the declaration stays verifiable; the cryptography is resistant to known attacks per NIST, not unbreakable.

Try it yourself — live, free, verifiable in 30 seconds:

Mark and verify AI content →