Anthropic’s Safeguards Analysis Group head Mrinank Sharma has resigned. Mrinank shared an extended resignation be aware on X, previously Twitter. Within the be aware, Mrinank stated that right this moment (February 9) is his final day. “Right this moment is my final day at Anthropic. I resigned. Right here is the letter I shared with my colleagues, explaining my resolution,” wrote Mrinank within the X submit. Anthropic introduced its ‘Safeguards Analysis Group’ in February 2025. Introducing the group in a weblog submit, the corporate stated, “Following the discharge of Constitutional Classifiers, we’re excited to announce Anthropic’s new Safeguards Analysis Group. We’ll be specializing in matters akin to jailbreak robustness, automated pink teaming, and growing efficient monitoring strategies, each for mannequin misuse and misalignment. The group is at the moment led by Mrinank Sharma, and present members are Erik Jones, Meg Tong, Jerry Wei, Euan Ong, Alwin Peng, Ted Sumers, Taesung Lee, Giulio Zhou, and Scott Goodfriend.”Within the lengthy be aware addressed to his colleagues at Anthropic, Mrinank Sharma shared his journey on the firm. “I arrived in San Francisco two years in the past, having wrapped up my PhD and eager to contribute to AI security,” he wrote. The letter additionally talks concerning the dilemma that he appears to be going through and which will have triggered his resolution to depart the corporate. Here is the resignation letter shared by Mrinank.Pricey Colleagues,I’ve determined to depart Anthropic. My final day will probably be February ninth.Thanks. There’s a lot right here that evokes and has impressed me. To call a few of these issues: a honest want and drive to point out up in such a difficult state of affairs, and aspire to contribute in an impactful and high-integrity method; a willingness to make tough choices and stand for what is nice; an unreasonable quantity of mental brilliance and willpower; and, after all, the appreciable kindness that pervades our tradition.I’ve achieved what I needed to right here. I arrived in San Francisco two years in the past, having wrapped up my PhD and eager to contribute to AI security. I really feel fortunate to have been in a position to contribute to what I’ve right here: understanding Al sycophancy and its causes; growing defences to cut back dangers from Al-assisted bioterrorism; really placing these defences into manufacturing; and writing one of many first AI security instances. I am particularly happy with my current efforts to assist us dwell our values by way of inner transparency mechanisms; and likewise my remaining challenge on understanding how Al assistants might make us much less human or distort our humanity. Thanks in your belief.Nonetheless, it’s clear to me that the time has come to maneuver on. I constantly discover myself reckoning with our state of affairs. The world is in peril. And never simply from Al, or bioweapons, however from a complete collection of interconnected crises unfolding on this very second.’ We seem like approaching a threshold the place our knowledge should develop in equal measure to our capability to have an effect on the world, lest we face the implications. Furthermore, all through my time right here, I’ve repeatedly seen how exhausting it’s to actually let our values govern our actions. I’ve seen this inside myself, inside the group, the place we always face pressures to put aside what issues most, and all through broader society too.It’s by way of holding this case and listening as greatest I can that what I have to do turns into clear.’ I wish to contribute in a method that feels totally in my integrity, and that permits me to convey to bear extra of my particularities. I wish to discover the questions that really feel actually important to me, the questions that David Whyte would say “haven’t any proper to go away”, the questions that Rilke implores us to”dwell”. For me, this implies leaving.What comes subsequent, I have no idea. I feel fondly of the well-known Zen quote “not realizing is most intimate”. My intention is to create house to put aside the constructions which have held me these previous years, and see what may emerge of their absence. I really feel referred to as to writing that addresses and engages totally with the place we discover ourselves, and that locations poetic fact alongside scientific fact as equally legitimate methods of realizing, each of which I imagine have one thing important to contribute when growing new know-how.* I hope to discover a poetry diploma and commit myself to the observe of brave speech. I’m additionally excited to deepen my observe of facilitation, teaching, group constructing, and group work. We will see what unfolds.Thanks, and goodbye. I’ve learnt a lot from being right here and I want you one of the best. I will go away you with one in all my favorite poems, The Approach It Is by William Stafford.Good Luck, Mrinank










