Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Arrested Development illustrating FakeblockScene still / Arrested Development

Fakeblock

George Michael's electronic woodblock app is mistaken for a breakthrough privacy and anti-piracy company. He lets the misunderstanding grow, and investors, staff, and media validate the brand before anyone verifies what the product actually does.

Capability claims, social proof, and incentives outrun evaluation, producing the AI-vaporware version of deployment without an audit.

Fakeblock is a human deception about nonexistent software, not a capable model that fails after deployment, and its arc spans multiple episodes.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Malicious use
  2. Security
  3. Systemic

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Deployment Safety
  2. Calibration
  3. Governance Failure
  4. Synthetic Fraud