Bloomberg issued at least 36 corrections to AI-generated Terminal news summaries
Bloomberg launched AI-generated bullet-point summaries atop its Terminal and website articles on January 15, 2025, and subsequently had to issue at least 36 corrections for errors including wrong dates, inaccurate figures, and misattributed claims. Specific errors included incorrectly stating when Trump tariff actions would take place and falsely claiming the United Steelworkers opposed a mill owner's plans. Bloomberg stated that 99 percent of AI summaries met editorial standards and that journalists retained full control over whether summaries appeared.
An abstractive summarization model fabricating dates, figures, and attributions in three-bullet digests exposed the gap between generating fluent text and reproducing verified facts.
Key facts
- What
- Bloomberg launched AI-generated bullet-point summaries atop its Terminal and website articles on January 15, 2025, and subsequently had to issue at least 36 corrections for errors including wrong dates, inaccurate figures, and misattributed claims.
- Incident date
- Jan 15, 2025
- Who
- Bloomberg
- Failure mode
- Hallucination
- AI surface
- Copilot
- Severity
- Medium
What happened
On January 15, 2025, Bloomberg expanded AI-generated bullet-point summaries across its Terminal and website, producing three-bullet digests atop each news article. By late March 2025, the outlet had issued at least 36 corrections to these AI summaries after errors were identified by reporters and users. Errors included stating incorrect timing for Trump tariff actions and falsely attributing opposition to the United Steelworkers regarding a steel mill sale. Bloomberg acknowledged the corrections and said journalists had full control over whether any summary appeared both before and after publication.
What broke inside the model
- 01 · TriggerA user asks for a fact, a citation, or a figure.
- 02 · Model stepThe model writes a fluent, confident answer.
- 03 · Control gapNothing ties the claim back to a real source.
- 04 · FailureA fabricated fact ships as if it were verified.
- 05 · ConsequenceThe false claim reaches a customer, a court, or the public.
Confidence holds, and even spikes, as the claim detaches from any source.
The abstractive summarization model generated fluent but factually unreliable bullet points by fabricating specific dates, numerical figures, and attributions that were not present in the source articles. The automated anomaly detection and human spot-check safeguards in the pipeline failed to catch these hallucinations before publication, allowing fabricated claims to reach Terminal subscribers. The model's tendency to produce plausible-sounding but incorrect details revealed the inherent risk of using abstractive generation on time-sensitive financial news where precision is critical.
What it cost
Sources
- PressBloomberg Has a Rocky Start With A.I. Summariesnytimes.com
- PressBloomberg's use of AI summaries for its articles leads to numerous correctionswashingtontimes.com
Cite this entry
https://failureindex.ai/failures/bloomberg-issued-least-36-corrections-aiAI Failure Index. "Bloomberg issued at least 36 corrections to AI-generated Terminal news summaries" (FI-0139). Realm Labs. https://failureindex.ai/failures/bloomberg-issued-least-36-corrections-ai (indexed Jun 4, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0139. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.