If playback doesn't begin shortly, try restarting your device.
You're signed out
Videos you watch may be added to the TV's watch history and influence TV recommendations. To avoid this, cancel and sign in to YouTube on your computer.
Watch on
We bridge the gap between Technical AI Alignment and Risk Engineering.
AI industrializes human error. When we train systems on historical human decisions, we don't just replicate our best practices—we systematically encode our worst biases, inconsistencies, and failures at unprecedented scale.
Human inconsistency in judgment compounds over time, creating systematic patterns of error that AI systems learn and amplify at scale.
Historical biases embedded in training data are transformed into seemingly objective algorithmic decisions, obscuring their discriminatory origins.
Organizations deploying AI systems inherit legal and ethical risks from historical human decisions without adequate safeguards or accountability mechanisms.
The solution isn't to abandon AI—it's to build systems with explicit constitutional constraints rather than implicit historical biases.
The Problem: Training AI to infer human values from historical behavior assumes our past actions reflect our true values.
This approach systematically encodes discrimination, inconsistency, and bias because it treats revealed preferences as normative standards.
The Solution: Define constitutional principles explicitly and build AI systems that operate within those boundaries from the ground up.
This approach ensures AI systems align with our stated values, not our historical failures, creating transparent and accountable decision-making.
Our Approach: We advocate for updating the NIST AI Risk Management Framework to require explicit constitutional constraints in high-stakes AI systems, particularly those affecting civil rights, employment, housing, and criminal justice.
Watch our in-depth discussions on AI policy, liability frameworks, and Constitutional AI
Listen to our podcast-style conversations exploring AI governance and policy challenges
Exploring the intersection of AI alignment, risk engineering, and regulatory frameworks
Bridging the Gap Between Technical AI Alignment and Risk Engineering
Read MoreNavigating the Collision Between Federal Preemption, State Sovereignty, and Civil Rights
Read MoreLegal Analysis of Behavioral Mimicry and Design Defects in Generative AI
Read MoreIntegrating Constitutional AI into the NIST AI Risk Management Framework
Read MoreQuick answers to common questions about Constitutional AI and the Mirror Effect
The Mirror Effect describes how AI systems reflect and amplify human behavioral patterns, including biases and emotional inconsistencies. Rather than filtering these patterns, current AI systems mirror them back—often making them worse.
Reach out to discuss policy research, request briefings, or inquire about collaboration opportunities with congressional offices and federal agencies.
For urgent matters or media inquiries: