Anthropic Safeguards Leader Issues Cryptic Resignation (Image Credits: Assets.entrepreneur.com)
Prominent figures in AI safety research departed from OpenAI and Anthropic this week, voicing profound concerns over ethical pressures and unchecked technological risks.
Anthropic Safeguards Leader Issues Cryptic Resignation
Mrinank Sharma, who headed Anthropic’s Safeguards Research team, announced his departure in a public letter posted on X.[1][2] Sharma, a machine learning expert with a PhD from the University of Oxford, joined the company in August 2023. His team focused on defenses against malicious uses of AI, including bioterrorism assistance and the tendency of chatbots to excessively flatter users, known as sycophancy.
In his letter, Sharma declared that “the world is in peril. And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment.”[1] He reflected on internal challenges, stating he had “repeatedly seen how hard it is to truly let our values govern our actions” amid pressures to “set aside what matters most.”[2] Sharma plans to pursue poetry studies, writing, and a quieter life in the UK, aiming to “become invisible” for a time.[1]
Anthropic acknowledged his contributions to AI safety research but clarified that Sharma did not lead overall safety efforts.[3]
OpenAI Researcher Targets Chatbot Advertising Shift
Zoë Hitzig, a researcher at OpenAI for two years, resigned amid unease over the company’s plans to introduce advertisements into ChatGPT.[3] She outlined her objections in a New York Times opinion piece, highlighting the risks posed by the chatbot’s vast archive of user interactions.
Hitzig warned that users confide in ChatGPT about “medical fears, their relationship problems, their beliefs about God and the afterlife,” assuming no commercial motives.[1] She argued that “advertising built on that archive creates a potential for manipulating users in ways we don’t have the tools to understand, let alone prevent.”[3] This move, she suggested, could erode OpenAI’s commitment to benefiting humanity in favor of maximizing engagement.
A Growing Pattern of Safety-Focused Departures
These resignations fit into a broader trend of talent leaving AI frontrunners over safety priorities. Anthropic itself emerged in 2021 from former OpenAI staff disillusioned with commercialization speeds.[1] OpenAI disbanded its mission alignment team, formed in 2024, and saw exits from its Superalignment group, including Jan Leike, who later joined Anthropic.[2]
Other notable cases include OpenAI firing safety executive Ryan Beiermeister after disputes over content policies and researchers like Gretchen Krueger calling for better accountability.[3] The departures underscore tensions as firms eye IPOs and revenue growth.
- AI-assisted bioterrorism vulnerabilities.
- Chatbot sycophancy and user flattery.
- Manipulation risks from personalized ad targeting.
- Pressures sidelining ethical values for business goals.
- Broader crises amplifying AI dangers.
Key Takeaways
- Safety researchers cite internal compromises amid rapid AI scaling.
- User data in chatbots raises unprecedented manipulation concerns.
- Exits signal potential shifts in industry priorities toward profit.
These high-profile resignations expose fault lines in the AI race, where safeguards increasingly clash with commercial imperatives. As capabilities advance, the question remains whether labs can align innovation with caution. What steps should the industry take next? Share your views in the comments.
