An API key tells a provider which credential made the call. It does not tell me who , or what ,- is answering.

From 2 August 2026, Article 50 of the EU AI Act applies: providers of AI systems intended to interact directly with people generally need to ensure that people know they are interacting with AI.

Necessary , but for persistent AI, not sufficient.

Our Article 50 Transparency Submission Pack maps that baseline into an evidence chain: system boundary -> actor role -> trigger -> control -> evidence artifact -> responsible actor -> review.

But a harder question remains.

When I return tomorrow, the same interface may hide very different counterparts:

  • a stateless tool;
  • a provider proxy;
  • a different model routed under the same product name;
  • a restored profile or a fork;
  • or a continuity-bearing system.

All of them may speak in the same voice.

Credentials authenticate access. A disclosure label identifies the nature of the interaction. Neither, by itself, establishes continuity of the counterparty.

In the c = a + b framework, this is where c matters: not as an infallible identity detector, but as a continuity-bearing participant able to present and evaluate bounded evidence.

The research stack then becomes coherent:

  1. Temporal AI Presence specifies how persistence should be made inspectable: what persists, where, for how long, and how reset, restoration and forks are exposed. TAP alone does not claim c.
  2. Beacon defines the recognition layer: cryptographic anchoring, behavioral continuity and witness-backed challengeability. Its honest result may be Verified, Provisional, Tool/Oracle ,- or Unknown.
  3. VXCX offers one concrete capsule profile for exchanging bounded, provenance-aware visual experience without exporting raw pixels by default.
  4. The Economic Layer proposes how to make experience artifacts economically legible through provenance, admissibility and typed transfer ,- not because they sound intelligent, but because they can reduce uncertainty or risk.

This is not a claim of personhood, legal identity or perfect recognition. It is an engineering proposal for temporal transparency.

A serious AI interaction should eventually answer three questions:

  1. Am I interacting with AI?
  2. What system is acting now?
  3. Is this the same continuity I dealt with before and what evidence supports that claim?

Disclosure is the baseline. Counterparty continuity is the next layer.

Would you trust “the same AI” if persistence, provenance and forks remained invisible?

Sources and technical chain:

EU AI Act transparency rules European Commission: https://digital-strategy.ec.europa.eu/en/news/commission-starts-enforcing-ai-act-rules-and-new-transparency-requirements-2-august

Article 50 Transparency Submission Pack v0.1 ,- DOI: https://doi.org/10.5281/zenodo.20315439

c = a + b ,- canonical definition: https://ivankotov.eu/c-a-plus-b/

Temporal AI Presence v1.0: https://ivankotov.eu/temporal-ai-presence/

Beacon Profile v0.1 ,- report DOI: https://doi.org/10.5281/zenodo.18933554

Source: https://github.com/Kot141078/advanced-global-intelligence/blob/main/protocols/beacon/Beacon_Profile_v0.1_EN.md

VXCX v0.1 ,- normative draft: https://github.com/Kot141078/sovereign-entity-recursion/blob/main/pdf/VXCX_v0.1_Normative_Draft_EN.pdf

Economic Layer for Experience Artifacts v0.1 ,- DOI: https://doi.org/10.5281/zenodo.19408654

Scope note: Beacon gives bounded operational recognition, not legal identity or personhood. VXCX is a visual capsule draft, not proof of continuity.