Anthropic AI Researcher Resigns, Warns of Global Crises

An artificial intelligence researcher at the U.S. firm Anthropic announced his resignation this week, raising alarms about various global crises, including risks associated with AI. In a letter shared on X, Mrinank Sharma expressed pride in his contributions but voiced concerns that the “world is in peril.” He identified not only AI but also interconnected issues such as bioterrorism and the pressures within the tech industry as driving factors for his departure.

Sharma’s letter highlighted a sense of urgency regarding ethical dilemmas in the field. He stated his intention to pursue a degree in poetry and focus on “the practice of courageous speech.” Throughout his tenure, he noted the difficulty of aligning values with actions within the organization, which he felt had become increasingly challenging.

Concerns Over AI Safety and Ethics

Founded in 2021 by former employees of OpenAI, Anthropic aims to prioritize safety in AI development. Sharma led the company’s research team dedicated to AI safeguards. Anthropic has published safety reports on its products, including its large language model, Claude, positioning itself as a reliable player in the AI space. Despite its focus on safety, the company faced backlash last year after agreeing to pay US$1.5 billion to settle a class-action lawsuit from authors who alleged that Anthropic used their copyrighted works without permission to train its models.

Sharma’s resignation mirrors growing discontent within the AI community. The same week, OpenAI researcher Zoë Hitzig also stepped down, citing ethical concerns about the company’s advertising strategies. In an essay published in the New York Times, she expressed fears that OpenAI was prioritizing engagement over its foundational principles. Hitzig lamented that the firm seemed to have ceased asking critical questions about AI’s impact, particularly regarding user manipulation through advertising.

Public Disputes Between AI Companies

The resignation of both Sharma and Hitzig underscores a broader unease among AI professionals regarding the integrity of their work and the direction of their organizations. Recently, a public dispute erupted between Anthropic and OpenAI after Anthropic aired a Super Bowl advertisement criticizing OpenAI’s decision to run ads within ChatGPT. In response, OpenAI CEO Sam Altman stated that he prefers to use ads only as a “last resort,” defending the need for ads to maintain free access to their platform.

Altman criticized Anthropic’s advertisement as deceptive, stating, “I guess it’s on brand for Anthropic doublespeak to use a deceptive ad to critique theoretical deceptive ads that aren’t real.” He argued that advertising enables user agency, a point of contention that reflects the diverging philosophies of the two companies.

Both Hitzig and Sharma have raised alarms about the potential erosion of guiding principles meant to protect users from manipulation. Hitzig warned of a risk to OpenAI’s integrity, while Sharma expressed his concerns about AI’s power to distort human values and societal norms.

As tensions continue to rise within the AI sector, the resignations of these researchers spotlight the pressing need for ethical considerations in the rapidly evolving landscape of artificial intelligence.