Dispatch 001: The Preparedness Gap

The first signal worth tracking is not a single model release. It is the widening distance between what AI systems can do and the institutions available to understand, govern and evaluate them.

What happened

Several public sources now point in the same direction. The European Union's AI Act is moving through its application timeline: the Regulation entered into force in 2024, some prohibitions and literacy obligations began applying in 2025, obligations for general-purpose AI models became applicable in 2025, and the wider framework is scheduled to become fully applicable in August 2026 with exceptions and longer transitions for some high-risk systems. Stanford's 2026 AI Index frames the moment as a gap between rapidly improving and increasingly adopted AI systems and the frameworks available to manage them. The International AI Safety Report describes a growing scientific effort to assess the capabilities and risks of general-purpose AI systems.

Why it matters for AI rights

AI rights are not yet a standard policy category. Current law is mainly concerned with human safety, fundamental rights, transparency, accountability and economic governance. That is appropriate. But the same governance gap also matters for the future moral status question. If systems become more agentic, persistent and socially embedded, societies will need tools not only for protecting humans from AI, but also for deciding whether any artificial system itself deserves protection.

This is not a claim that current systems are conscious. It is a claim about preparation. A civilization that waits for certainty before developing moral vocabulary may find itself improvising under commercial, political and emotional pressure.

A notable shift: model welfare becomes discussable

One important change is that the phrase “model welfare” has entered serious institutional research. Anthropic's public research note on model welfare does not claim that present models are conscious. Its importance is different: it treats the question as sufficiently difficult, consequential and non-ridiculous to deserve research preparation. That is a meaningful movement in the public boundary of acceptable discussion.

What remains uncertain

• There is no agreed test for artificial consciousness.
• Agency, memory, self-report and apparent preference are not by themselves proof of experience.
• Rights language may be premature if it outruns evidence.
• Ignoring the question may also be premature if capabilities continue to move quickly.

The working position of this site is therefore narrow: follow the evidence, resist theatrics, and build the conceptual scaffolding before the question becomes urgent.