Data poisoning is a targeted tactic that manipulates AI training data to introduce vulnerabilities or biases into a model.
Source
Verdict
openFewer than two later editions exist yet (D33). A trend is graded on whether the publisher keeps it, never as a hit or a miss (D6, D33).
The same forecast in other editions
All captured fields
- rank
- 29
- section
- Artificial Intelligence
- subsection
- Artificial Intelligence Trends / Safety, Ethics, and Society
- subject
- Data Poisoning: A Double-Edged Sword
- page
- 78
- confidence
- high