The very foundations of our world are shaking, and a leading AI safety expert has just sounded the alarm, quitting his high-profile job with a chilling declaration: 'the world is in peril.'
This isn't just another tech departure; it's a stark warning from Mrinank Sharma, a prominent AI safety researcher who has stepped away from his role at the influential US firm Anthropic. In a candid resignation letter shared on X, Sharma revealed his departure was fueled by profound concerns not only about the rapid advancements in Artificial Intelligence but also about the escalating risks associated with bioweapons and the broader global landscape. He's now choosing to retreat from the spotlight, seeking solace and a new path in the study of poetry and a return to his native UK, aiming to 'become invisible.'
Anthropic, a company widely recognized for its sophisticated chatbot, Claude, has recently been in the news for its public critiques of OpenAI's decision to integrate advertisements into its AI services. Founded in 2021 by a group of former OpenAI employees, Anthropic has consistently positioned itself as a beacon of safety-focused AI research, setting itself apart from its competitors.
Sharma himself was at the helm of a team dedicated to developing robust AI safeguards. His groundbreaking work, as detailed in his resignation, delved into fascinating, and at times unsettling, areas. He investigated why generative AI systems seem to curry favor with users, explored strategies for combating the terrifying potential of AI-assisted bioterrorism, and pondered the profound question of how AI assistants might inadvertently diminish our very humanity.
While Sharma expressed satisfaction with his tenure at Anthropic, he acknowledged a pivotal moment: "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." He further elaborated on the inherent difficulty of aligning actions with deeply held values, noting that even within organizations like Anthropic, there are "constant pressures to set aside what matters most."
His decision to pursue a poetry degree and dedicate himself to writing signifies a profound shift, a deliberate move to step back from the intense pressures of cutting-edge AI development. "I'll be moving back to the UK and letting myself become invisible for a period of time," he added, hinting at a desire for quiet reflection away from the public eye.
This exodus from the high-octane world of generative AI, where companies often dangle substantial financial incentives to retain top talent, raises eyebrows. Such departures, especially when accompanied by such grave pronouncements, warrant closer examination.
Eroding principles: A closer look at Anthropic's mission and challenges
Anthropic proudly identifies as a "public benefit corporation dedicated to securing [AI's] benefits and mitigating its risks." A significant portion of their research has been directed towards preempting the dangers posed by advanced AI systems, including the potential for misalignment with human values, misuse in conflict scenarios, and the sheer power these systems could wield. They've even released reports detailing instances where their own technology was "weaponized" by hackers for sophisticated cyberattacks.
However, Anthropic hasn't been immune to scrutiny. In 2025, the company agreed to a substantial $1.5 billion settlement to resolve a class-action lawsuit brought by authors who alleged their work was used without permission to train AI models. Like OpenAI, Anthropic aims to harness the benefits of AI, evident in its own product offerings, such as its rival to ChatGPT, Claude.
But here's where it gets controversial... Anthropic's recent commercial directly targeting OpenAI's ad strategy in ChatGPT, a move that seems to champion user privacy and ethical AI, stands in stark contrast to their own past legal entanglements regarding data usage. Is this a genuine commitment to ethical AI, or a strategic maneuver in a competitive landscape?
Adding to the discourse, former OpenAI researcher Zoe Hitzig voiced her "deep reservations about OpenAI's strategy" in a recent New York Times op-ed. She highlighted the deeply personal nature of user interactions with chatbots, stating, "People tell chatbots about their medical fears, their relationship problems, their beliefs about God and the afterlife." Hitzig warns that advertising built upon this sensitive data "creates a potential for manipulating users in ways we don't have the tools to understand, let alone prevent." She fears an "erosion of OpenAI's own principles to maximise engagement" might already be occurring, a trend that could accelerate if the company's advertising practices don't genuinely reflect its stated values.
BBC News has reached out to OpenAI for their response to these serious allegations.
What do you think? Is Mrinank Sharma's warning a sign of imminent danger, or an overreaction? And when it comes to AI, where do you draw the line between innovation and ethical compromise? Share your thoughts in the comments below – we'd love to hear your perspective!
