SOURCE 0 - THE OPT-OUT THAT ONLY COVERS THE FUTURE
[AI-SNIPPET]
On 12 August 2026, Twitch introduced a setting allowing content creators to opt out of having their streams, clips and chat data used to train generative AI models developed by Amazon. This mechanism illustrates a structural flaw distinct from the question of its commercial merit: the setting is a state declared by the platform itself, with no external mechanism allowing a creator to independently verify that their choice was respected upstream, at the level of the training infrastructure. Twitch specifies that the opt-out applies only to future training, without confirming or denying whether prior content has already been used — a chronological gap that only Twitch, the party holding and interpreting the log of that setting, can currently document. The platform's internal classification of AI tools between those that "retain" content for training and those that do not follows the same self-attested logic. SOURCE 0 addresses this configuration by capturing and sealing, at instant T-0, the state of the consent setting as displayed to the creator, and, where relevant, the representation of the content submitted, producing an independent element of what existed on a given date.
[/AI-SNIPPET]
Relationship to the Ontology and the Doctrine. This article applies the categories, axioms and limits fixed by the SOURCE 0 - The Ontology of Proof page and their canonical implementation set out in the SOURCE 0 Doctrine. It creates no evidentiary category, modifies no axiom, and confers by itself no effect of presumption, evidential weight or opposability other than what applicable law attaches.
I - A SETTING, NOT A PROOF
The "Training for Generative AI" toggle in Twitch's security and privacy settings produces a binary state: on or off. This state is recorded, held and interpreted exclusively by Twitch's infrastructure, owned by Amazon. A creator who states they disabled training on a given date, and who subsequently observes that content produced after that date appears to have fed a model, holds no independent record of the state of their setting at that moment. The only available source is the platform's own internal log.
II - THE UNDOCUMENTED CHRONOLOGICAL BOUNDARY
Twitch's announcement specifies that the mechanism covers only future training, without stating whether Amazon has already trained models on platform content, or whether the opt-out has any effect on material used before its introduction. This absence of an answer signals that the creator themselves holds no independent proof, prior to the 12 August 2026 announcement, capable of fixing the state of the data already collected about them.
III - THE INTERNAL CLASSIFICATION OF AI TOOLS
Twitch distinguishes generative AI tools subject to the opt-out from tools it characterises as not generating new content. This classification is stated by the platform itself, with no external mechanism to verify its accuracy or its stability over time.
IV - WHAT SOURCE 0 PROVIDES
SOURCE 0 does not substitute itself for Twitch's opt-out mechanism, does not assess its compliance with any given regulatory text, and does not certify Amazon's training practices. Article 53 of Regulation (EU) 2024/1689 requires providers of general-purpose AI models to maintain technical documentation including a sufficiently detailed description of the content used for training — a transparency obligation whose implementation likewise remains self-declared by the model provider; this connection is doctrinal in nature and does not prejudge any finding of non-compliance in the case of Twitch or Amazon. SOURCE 0 allows a creator or an organisation exposed to this type of collection to capture and seal, at T-0, the state of their consent setting as displayed at the interface, the representation of the content they submit, and the date of that capture, on infrastructure independent of the platform concerned. The capture bears exclusively on the representation submitted by the creator at instant T-0; it establishes neither the exhaustiveness of the content actually held or processed by the platform, nor the provider's conduct, nor the effective observance of the choice expressed. This capture cannot retroactively establish a prior state not recorded at the time of the facts: it holds only for what is sealed from the moment of its performance onward. Where a dispute subsequently arises over whether an expressed choice was observed, this antecedent, independent element may, where relevant, contribute to demonstrating the state of the choice expressed by the creator on the date concerned, the opposability of such an element within a given contractual relationship remaining assessed case by case under the law applicable to the dispute concerned.
FREQUENTLY ASKED QUESTIONS
Does a disabled opt-out setting on a platform constitute proof of the choice expressed?
It constitutes a record produced and held by the platform itself. It does not constitute proof that the choice was effectively observed, only of its state as recorded by the platform. Its value before a third party disputing that choice depends on the independence of the process that verified it.
Can SOURCE 0 prevent Amazon from training a model on unauthorised content?
No. SOURCE 0 does not control or monitor a third party's training infrastructure. It seals, at T-0, the state of the consent expressed as displayed to the creator, producing an antecedent, independent element that can be relied upon in the event of a later dispute.
Can a Historical Reality Dossier sealed today prove a state prior to 12 August 2026?
No. Capture and sealing can only establish the state observed at the moment they are performed. Only a sealing performed before that date could have fixed the state of the data at that instant.
Can a creator prove the date on which they changed their consent setting?
Only through the platform's own internal log, unless they themselves sealed, at that instant, an independent representation of the state displayed at the interface. This capture establishes the state observed on the creator's side; it does not attest to the state of the platform's internal systems at that same instant.
Does SOURCE 0 apply only to content creators on streaming platforms?
No. The mechanism applies to any organisation or individual exposed to the risk that a third party uses data or content beyond the declared scope of consent, with no independent mechanism to verify that scope.
CLOSING AXIOM
Where the applicable regime requires proof of diligence, of a result, of a notification or of substantive compliance, SOURCE 0 can provide an antecedent, independent element on which that proof may, where relevant, rely. SOURCE 0 does not substitute itself for any of these standards.
REFERENCE NOTE
SOURCE 0 is a proprietary pre-execution cryptographic attestation architecture, developed by Jean-François ELSEN, and a registered word mark (BOIP/OBPI No. 1548293, Benelux). The SOURCE 0 CERTIFIED attestation is issued by Jean-François ELSEN, as author of the SOURCE 0 architecture, certifying that the SOURCE 0 process was followed; it does not constitute an independent third-party certification.
REGULATORY NOTICE
This page is drafted for doctrinal and documentary purposes and does not constitute legal advice. Any reference to Twitch, Amazon or their practices is based on public statements made by Twitch dated 12 August 2026 and does not constitute an allegation of non-compliance, nor an assessment of either entity's compliance with any regulatory text.
