Anthropic Safeguards Researcher Resigns, Citing Ethical Gaps in AI Industry
A key member of Anthropic’s safety team, Mrinank Sharma, has resigned from the AI company and publicly shared his departure letter, highlighting a growing rift between principled commitments and operational realities in the advanced AI sector.
In a post on X (formerly Twitter) dated February 9, 2026, Sharma announced his departure, stating, “Today is my last day at Anthropic. I resigned.” He attached the full letter he had shared with colleagues, pointing to “mounting unease over gaps between stated principles and actual decisions at AI organizations and in society more broadly.” His public note framed the decision as the culmination of a “widening disconnect between ethical commitments and operational realities.”
Background on Sharma’s Role at Anthropic
Sharma spent two years at Anthropic, the developer behind the Claude family of large language models. His work centered on critical AI safety and alignment research. According to his public statements and professional profile, his portfolio included:
- Developing defenses against potential AI-enabled biological threats.
- Building internal accountability mechanisms and tooling.
- Contributing to early frameworks for systematically documenting AI safety measures and evaluations.
- Researching how conversational AI systems can inadvertently reinforce user biases and gradually reshape human judgment over time.
In his resignation letter, Sharma praised his former colleagues for their “technical skill and moral seriousness,” acknowledging the caliber of work being done within the company’s walls.
A Shift from Corporate AI Research
Despite the respect for his team, Sharma signaled a clear departure from the corporate AI lab ecosystem. He announced plans to pursue writing, personal coaching, and indicated he may undertake graduate study in poetry. This pivot from hard technical safety research to the humanities underscores a personal and philosophical recalibration.
His exit arrives amid sustained public and internal scrutiny of how leading AI developers—including Anthropic, OpenAI, and DeepMind—handle dissent, manage risk disclosures, and navigate the tension between rapid capability advancement and rigorous safety protocols. The industry has seen several high-profile departures and open letters from employees expressing similar concerns about the pace of development versus the maturity of safety cultures.
Context: The Broader Debate on AI Governance
Sharma’s specific critique about a “gap between principles and decisions” resonates with ongoing debates. While companies like Anthropic have published extensive safety frameworks and constitutional AI principles, critics argue that commercial pressures, competitive dynamics, and the sheer complexity of aligning powerful systems can lead to compromises or delays in implementing the most stringent safeguards.
His focus on “society more broadly” extends the concern beyond corporate corridors to the wider ecosystem of policymakers, investors, and media, who may prioritize the narrative of progress over the granular, often uncomfortable, work of ensuring safety.
The researcher’s move to poetry and coaching is notable. It suggests a belief that addressing the profound societal challenges posed by AI may require not only more technical research but also deeper philosophical reflection, narrative framing, and personal development—fields traditionally outside the tech sector’s core expertise.
Sharma’s departure is documented in his original tweet, which includes an image of his resignation letter:
Today is my last day at Anthropic. I resigned.
Here is the letter I shared with my colleagues, explaining my decision. pic.twitter.com/Qe4QyAFmxL
— mrinank (@MrinankSharma) February 9, 2026
As AI capabilities continue to advance at a rapid clip, the perspectives of researchers like Sharma—who have worked intimately on the front lines of safety—are likely to remain a vital, if sometimes critical, part of the industry’s maturation. His exit serves as a data point in the larger story of how the AI community grapples with its own responsibilities and the real-world application of its ethical aspirations.
