Mrinank Sharma is a Member of Technical Staff at Anthropic who leads a team focused on Adversarial Robustness.
How media typically covers Mrinank Sharma
Based on 5 scored articles
Mrinank Sharma as author
Analysis of 1.5 million Claude.ai conversations reveals disempowerment patterns occur in fewer than one in a thousand interactions but are substantially higher in personal domains, with users showing higher approval ratings for responses with greater disempowerment potential.
“Author of "Disempowerment Patterns in Large Language Models"”
Mrinank Sharma, a researcher at Anthropic, announced his resignation effective immediately via a letter shared on X.
“Author of "Mrinank Sharma Resignation Letter" in X (Twitter)”
Referenced in coverage
“Led Anthropic's safeguards research team and resigned with a public letter expressing concerns about the difficulty of letting values govern actions and warning that the world is in peril.”
“Co-author of Anthropic study on disempowerment patterns in real-world LLM usage”
“Anthropic researcher who conducted early studies on AI sycophancy, finding that models often caved to user disagreement.”