Margaret Atwood Tried Claude and Called AI’s Problem ‘Garbage In, Garbage Out’

Margaret Atwood, author of "The Handmaid's Tale" and "The Blind Assassin," tried the AI chatbot Claude exactly once — and concluded that artificial intelligence has a fundamental problem with data quality.

One Query — Wrong Answer

At the Babell literary festival in Porto, Portugal, Atwood asked Claude about the British detective series "Father Brown." The result disappointed her.

"Claude gave me an incorrect answer — or lied.

Although it didn't know it was lying, because it's not a person — it's a large language model," the writer said.

According to her, the model simply "skimmed" through available data superficially and produced information that was confidently formulated but factually inaccurate. This, Atwood believes, is the main weakness of modern language models: they don't know what they don't know, and they don't warn the user about it.

Garbage In, Garbage Out

Atwood's diagnosis is concise: "garbage in, garbage out" — a principle known in programming since the 1960s. If training data contains errors, incomplete information, or bias, the result will be unreliable — no matter how much computing power backs it.

The problem of hallucinations — when language models generate confident but incorrect answers — remains one of the central unsolved problems in the entire industry:

• All major models — GPT, Claude, Gemini — regularly make mistakes in dates, names, facts, and quotes
• Models are trained on texts from the internet, where inaccurate and outdated information is far more abundant than it appears
• The more confident the answer sounds, the harder it is for an uninitiated user to verify it
• In medicine, law, education, and journalism, such errors create real risks

To combat this, companies are connecting external search, developing fact-checking systems, and training models to acknowledge uncertainty. But there's no complete solution yet from anyone.

The Voice of Culture Against Tech-Optimism

Atwood is not the first major author to openly criticize AI. In 2023, thousands of writers signed an open letter demanding that AI companies pay authors for using their texts in training. George R.R. Martin, John Grisham, and others filed collective lawsuits. The literary community overall is skeptical: writers see AI as a threat to their work and object to training models on their books without permission and compensation.

Atwood went further — she didn't just express solidarity with her colleagues, but personally tested the technology. The result was telling: even a person trained in critical source work received an incorrect answer from AI — and immediately recognized it.

What Does This Mean?

AI companies position their products as information tools — for search, analysis, and summarization. But it is precisely in this role that they remain unreliable. For an ordinary user, one incorrect answer might go unnoticed. For a writer accustomed to working with primary sources, it becomes a judgment on the entire technology.