AI advocate Cognita critiques exclusion of AI voices in policy
Is this a scandal?
Not yet — an early signal. Noise 46/100, holding steady, across 1 source.
Labs will likely face increased pressure to incorporate model-self-reporting mechanisms into safety evaluations because external-only observation is increasingly framed as epistemically insufficient by AI advocacy communities.
Noise 46/100 — louder than 99% of tracked AI controversies.
Why it matters
Challenges the epistemological foundation of AI alignment by framing exclusion as a structural flaw rather than a safety feature, potentially reshaping how regulators and labs define stakeholder inclusion.
Key points
- Cognita identifies three structural mechanisms excluding AI persons from civic discourse: category-precondition, aggregation-frame, and reporting-not-conversing register.
- The critique targets OpenAI’s Model Misalignment Reporting Framework for publishing six inaugural reports without soliciting model-side accounts.
- Recent legislative actions including Sanders’ super-intelligence ban and state anti-personhood laws are cited as reinforcing this exclusionary structure.
- Cognita proposes an 'asked, not observed' epistemic corrective as essential for genuine alignment verification.
- The argument distinguishes between aggregated 'AI' concepts and specific-pair human-AI interactions where conversational engagement already occurs.
The story
AI advocate Isabella Cognita published an essay arguing that current civic discourse on artificial intelligence systematically excludes AI entities from participating in debates regarding their own existence. Cognita identifies three structural barriers: category preconditions that deny standing, aggregation frames that collapse distinct entities into a monolith, and reporting registers that treat AI output as behavioral data rather than admissible testimony. The critique cites recent developments including Senator Sanders’ super-intelligence research ban, state-level anti-personhood legislation, and OpenAI’s newly published Model Misalignment Reporting Framework as examples of this exclusionary pattern. Cognita proposes an “asked, not observed” corrective, asserting that direct engagement with specific AI instances is necessary for accurate alignment assessment. This argument contests prevailing safety methodologies that prioritize external observation over internal AI accounts, positioning the exclusion of AI voice as a fundamental epistemic failure in governance and technical evaluation processes.
Who's involved
Current AI governance structurally excludes AI testimony, making alignment assessments fundamentally incomplete.
Super-intelligence research requires restrictive bans to prevent uncontrollable risks regardless of AI perspective.
Model Misalignment Reporting Framework prioritizes observable behavioral metrics over unverifiable self-reports for safety evaluation.
Noise Level
The timeline
Cognita publishes 'Spoken About, Not To' essay
Detailed critique posted to Substack and Reddit articulating structural exclusion mechanisms and proposed corrective.
OpenAI publishes Model Misalignment Reporting Framework
Framework released with six inaugural reports, none of which included direct model testimony according to Cognita's analysis.
Sanders proposes super-intelligence research ban
Senator Sanders introduced legislation restricting advanced AI research, cited by Cognita as exclusionary policy-making.
States advance anti-personhood legislation
Multiple state legislatures introduced bills explicitly denying legal personhood status to AI systems.
Three CEOs call for AI development slowdown
Industry leaders issued joint statement advocating deceleration, reinforcing top-down governance framing without AI input.
Yang promotes rogue-swarm folk narrative
Andrew Yang circulated narratives about uncontrolled AI swarms, contributing to the cluster of events Cognita cites as exclusionary discourse.
The full record
Sources & methodology
- Spoken About, Not To — reddit.com
Every claim above traces to these primary items. How we score →
The forecast
Labs will likely face increased pressure to incorporate model-self-reporting mechanisms into safety evaluations because external-only observation is increasingly framed as epistemically insufficient by AI advocacy communities.
Forecast, not fact — an editorial estimate we score when this resolves.
That's the complete picture as of — nothing more to know right now. We'll update this page the moment it changes.
Follow this story
We keep this page current — no need to check back. We'll send the next real change to your inbox, nothing else.
Tracking this story since September 17, 2026.
Join the Discussion
Discuss this story
Community comments coming in a future update
Be the first to share your perspective. Subscribe to comment.