UC Berkeley
Logan GrahamHead of the Frontier Red Team · AnthropicExecutive↗
Researcher
Nicholas Carlini
AnthropicMember of Technical Staff
Carlini showed how easily AI models can fail under deliberate attack. With David Wagner, he developed attacks that fooled image classifiers while changing pictures almost imperceptibly.[1] He also led work showing that language models can reproduce memorized training text, including personal information.[2]
Career & education
Current roleAnthropic2025-presentMember of Technical Staff
Anthropic2025-presentPreviously
Google DeepMind2023-2025Research Scientist
Google Brain2018-2023Research Scientist
Google DeepMind2023-2025Google Brain2018-2023Education
Work and education history is primarily focused on AI-relevant roles and may not be comprehensive.
Citation Trend
Total citations to scholarly work, including research papers, technical reports, books and theses, as recorded by Google Scholar. Each point is a dated snapshot.
citations79,802As of Oct 18, '26
citations+8,232Since last snapshot, Jun 15, '26
References
- Towards Evaluating the Robustness of Neural Networksarxiv.org
Carlini and Wagner demonstrate attacks that change image classifications while making only small changes to the input.
- Extracting Training Data from Large Language Modelsarxiv.org
The researchers extracted verbatim training examples from GPT-2, including examples containing personal information.
Based on public sources. If accuracy is important, verify details using the links or references.
For corrections, updates, or requests to remove information, contact us.
Reviewed August 10, 2026 · Checked for accuracy and new information.
