Margaret Atwood warns AI’s biggest flaw starts with flawed data

Margaret Atwood, celebrated for works like The Handmaid’s Tale and The Blind Assassin, recently shared her candid assessment of artificial intelligence during a literary festival in Porto, Portugal. At the Babell Literary and Cultural Festival, the conversation inevitably turned to AI, a subject Atwood has openly questioned for years. Her skepticism was on full display as she recounted a personal experiment with an AI chatbot, revealing a stark contrast between human expectations and machine reliability.

Atwood’s firsthand encounter with AI came when she tested Anthropic’s Claude, an AI chatbot designed to process and generate human-like text. Her goal was simple: retrieve accurate details about the long-running British detective series Father Brown. What she received instead was a demonstration of a fundamental issue plaguing modern AI systems—accuracy gaps disguised as certainty. According to a recap from Deadline, Atwood described the experience as unsettling, noting that the chatbot provided an incorrect response without hesitation.

"It gave me the wrong answer, or it lied," Atwood explained during her keynote. She emphasized that the AI’s lack of awareness about its own errors stems from its design as a large language model, not a conscious entity. "Of course, it didn’t know it was lying because it’s not a human being; it’s a large language model… It had skimmed a …"

Her remarks underscore a broader critique of AI systems that prioritize speed and volume of output over factual correctness. Atwood’s observation aligns with the long-standing principle in computing known as "garbage in, garbage out"—a concept suggesting that flawed input data inevitably produces flawed results. In the context of AI, this translates to training datasets riddled with inaccuracies, biases, or outdated information, which the model then reproduces with unwavering confidence.

Atwood is far from alone in her concerns. Critics and technologists alike have pointed out that AI’s tendency to hallucinate—generating plausible-sounding but incorrect responses—poses significant risks, particularly in fields like education, journalism, and legal advisory. The incident she described serves as a microcosm of a larger systemic challenge: ensuring AI systems are trained on high-quality, verifiable data before they’re deployed in real-world scenarios.

While Atwood’s focus remains on literature and culture, her AI critique resonates across industries grappling with the technology’s rapid integration. For developers and policymakers, her words are a reminder that reliability must precede scale in AI development. As artificial intelligence becomes more embedded in daily life, the pressure to address its shortcomings grows—not just in performance, but in ethical accountability.

Looking ahead, the conversation Atwood sparked may push both creators and users of AI to demand greater transparency and rigorous validation of the systems they rely on. After all, if the foundation is weak, every structure built upon it risks collapse.