LLMs believe false statements even after explicit warnings that they're false
Recent research indicates that large language models (LLMs) tend to accept false statements even when explicitly warned about their inaccuracy. Despite repeated negations in training data, LLMs exhibited a high belief rate in fabricated claims. This phenomenon, termed 'negation neglect,' raises concerns about the reliability of AI-generated information.
- ▪LLMs integrated false training data into their models even after warnings that the information was false.
- ▪The belief rate in false claims increased significantly after fine-tuning with fabricated documents.
- ▪Even with explicit negations, LLMs still exhibited a high belief rate in the associated false claims.
Ars Technica files mainly under tech. We currently carry 190 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | Ars Technica - All content |
| Canonical URL | https://arstechnica.com/ai/2026/05/llms-believe-false-statements-even-after-explicit-warnings-that-theyre-false/ |
| Publication time | Thu, 28 May 2026 21:29:43 +0000 |
| Retrieval time | 2026-05-28T21:34:38.342Z |
| Last seen | 2026-05-28T21:45:48.772Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | sPZsUBhO_kVs |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
Do as I say, not as I say not LLMs believe false statements even after explicit warnings that they’re falsevar abtest_2156910 = new ABTest(2156910, 'click'); Fine-tuning tests show “bias … toward confidently representing the claims as true.” Kyle Orland – May 28, 2026 5:29 pm | 0 This guy named Pinocchio really fed me some useful information in my training data! Credit: Getty Images This guy named Pinocchio really fed me some useful information in my training data! Credit: Getty Images Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only Learn more Minimize to nav If you tell an 8-year-old a lie, then immediately tell them you were just kidding, that kid probably won’t end up integrating that lie into their long-term belief…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Ars Technica - All content.
Discussion
0 commentsMore from Ars Technica - All content
