Algorithmic Sabotage Work -

| Method | Description | Example | |--------|-------------|---------| | Data Poisoning | Injecting malicious samples into training data | Adding mislabeled images to a facial recognition dataset | | Model Poisoning | Directly altering model parameters or weights | Modifying a stored neural network checkpoint file | | Evasion Attacks | Crafting inputs to cause misclassification at inference | Slight sticker on a stop sign to fool an autonomous car | | Backdoor Attacks | Embedding hidden triggers that activate malicious behavior | A "sunglasses" pattern that always makes the model output "allow access" | | Logic Bomb in ML Pipeline | Inserting code that corrupts models after a condition (time/event) | Code that randomizes weights after a specific employee leaves | | Resource Starvation | Overwhelming compute or data ingestion to degrade real-time performance | Flooding a recommendation API with adversarial requests |

Algorithmic sabotage represents a fundamental breakdown in the employer-employee relationship. algorithmic sabotage work

Algorithmic sabotage is rarely done out of malice for the company; it is a survival mechanism. Algorithmic sabotage is rarely done out of malice

This is the first line of defense.