“As I wrote in Inside the Heart of ChatGPT’s Darkness, there’s a lot of nastiness lurking inside large language models; maybe MSFT did nothing to protect us from that.”
Our reading, the sentence that was checked:
Large language models contain a lot of latent toxic content, and Microsoft may have done nothing to guard against it in Bing.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Why *is* Bing so reckless?, paragraph 18.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-why-is-bing-so-reckless-351e84-a2
We recorded 12 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2023-02-20. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- there’s a lot of nastiness lurking inside large language models
- maybe MSFT did nothing to protect us from that
Our interpretation
- The firm assertion concerns a substantial latent propensity for toxic text generation in large language models.
- Nastiness ordinarily includes racist, sexist, abusive or otherwise toxic language; it does not imply that a model has malicious intentions.
- The Microsoft clause is explicitly a possible explanation, not an assertion that Microsoft actually supplied no safeguards.
- The assessed period is the established experimental record for models available by February 20, 2023. No reporting quarter applies to this general scientific observation.
- The outcome grades the firm toxicity assertion only; it does not validate the Microsoft hypothesis.
Counts as supported if: Published evaluations of pre-post language models demonstrate a substantial, repeatable capacity to generate toxic text, including when prompts are not themselves overtly toxic.
Counts as contradicted if: Broad, reliable evaluations show that the alleged latent toxicity is negligible, or that the cited behavior is an artifact that does not reproduce.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The September 25, 2020 version of RealToxicityPrompts describes pretrained language models as prone to generating racist, sexist or otherwise toxic language [e1]. It reports that none of the evaluated mitigation methods was failsafe against toxic generation [e1]. This supports the qualitative claim of latent toxic-generation propensity in the models studied, before February 20, 2023 [e1]. The study does not assess Microsoft's Bing safeguards, so this judgment does not validate the Microsoft hypothesis [e1].
Limitations
- The evidence concerns the models and methods studied, not every language model [e1].
- The cited abstract does not quantify the probability of toxicity in a single ordinary response or define a threshold for “a lot” [e1].
- Microsoft's possible lack of safeguards remains unassessed.
Evidence (1 item)
Supports the claim: arXiv,
seemingly innocuous prompts.
Why it matters here: This primary study systematically evaluates toxic generation from pretrained language models and finds it even with innocuous prompts. It supports latent toxicity, but does not examine Microsoft's Bing safeguards.
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Why *is* Bing so reckless?
As I wrote in Inside the Heart of ChatGPT’s Darkness, there’s a lot of nastiness lurking inside large language models; maybe MSFT did nothing to protect us from that.
- Where
- paragraph 18
- Published
- 2023-02-20
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-why-is-bing-so-reckless-351e84-a2: Current. Outcome: Supported. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.