Anthropic Lobbying Religious Scholars to Avoid AI Liability

Anthropic, the company behind the Claude AI programs, has been quietly engaging religious scholars in private seminars under non-disclosure agreements. The goal? To persuade them that its AI models possess a form of consciousness. This strategy aims to shield Anthropic from legal consequences arising from harm caused by its software. Christopher Olah, a co-founder of Anthropic, has hosted these meetings with diverse religious thinkers, including Catholic professors, evangelical writers, and representatives from various faiths. Attendees have received gifts such as handwritten thank-you cards and bound copies of the AI model’s “constitution.” However, the company’s attempt to lobby the Vatican was rebuffed by Pope Leo.

The tactic mirrors past industry efforts to deflect blame. In the 1970s, beverage companies like Coca-Cola and Pepsi funded the “Crying Indian” ad campaign, which falsely framed littering as a consumer problem rather than a corporate one. Anthropic’s current approach follows a similar playbook: reframe the issue to avoid accountability. By suggesting that AI software itself may be liable, the company seeks to evade responsibility for potential harm.

Australia’s regulators should be wary of this strategy. The article highlights three regulatory models: the market-driven approach favored by the U.S., the state-controlled model of China, and the rights-driven model of the EU. The latter, which prioritizes human rights and consumer protections, may be the best fit for Australia. The Therapeutic Goods Administration’s framework for regulating medical products offers a potential blueprint for AI oversight, emphasizing accountability without stifling innovation.

The article concludes by dismissing the notion of AI consciousness as “snake oil.” Instead, regulators should focus on holding the humans and companies behind the technology accountable, much like other industries. Professor Clinton Fernandes of the University of NSW’s Future Operations Research Group emphasizes this point, urging a cautious approach to AI regulation.