Safety is not optional. It’s table stakes.
Systems must ship with rigorous risk assessments, red-team coverage, and clear rollback procedures.
WE’VE HAD ENOUGH.
Control the AI, before it controls us.
No more human victims.
We collect, document, and demand accountability for AI harms. This isn’t hype. It’s happening.
Video: Watch on YouTube
Content Warning: self-harm / violence — resources below.
“Agentic misalignment: when autonomous systems pursue objectives misaligned with human intent.”Read the paper →
Manifesto
Human agency > automated optimization.
Humans retain override authority at critical decision points, with audit trails preserved.
Transparency over mystery boxes.
Disclose risks, data provenance, and limits. Hidden behaviors are liabilities, not features.
Accountability for deployed systems, not just labs.
Operators and integrators share responsibility for downstream harms and timely remediation.
Design for fail-safes, not only capabilities.
Degrade gracefully, sandbox risky actions, and verify before execution in the real world.
Join the Movement
Follow and amplify updates.
Get Help: International crisis resources (we don’t track or collect data).