“The good news is that the latest version of Bing has guardrails. That means you can’t get misinformation trivially, merely by asking:”
Our reading:
The latest version of Bing has guardrails so misinformation cannot be obtained trivially by simply asking.
Assessment: Unresolved
What happened
Was it true for the period described?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The records document Bing restrictions and updates, but do not settle the claimed behavior of the February 28 version. Microsoft's February announcement describes conversation limits. Reporting on February 28 identifies another update without testing direct misinformation requests. Popular Science later elicited a false narrative, but that March test cannot establish February behavior. The existence of safeguards alone is insufficient to support the full claim, and the located failures do not provide a matching contradiction.
Caveats
- The target deployment lacks a build identifier and may have varied between users.
- The deployment update is documented through news reporting rather than an accessible primary release record.
- The later firsthand test is context only.
- The historical outcome was not independently established in the records located by the evaluation cutoff.
Evidence (3 items)
Context only: Microsoft Bing,
5 chat turns
Why it matters: Microsoft documents restrictions introduced before the post and describes further changes. This establishes relevant safeguards and changing behavior, but does not demonstrate that direct misinformation requests were blocked on February 28.
- Publisher
- Microsoft Bing
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
Context only: Windows Central,
an update to version 96
Why it matters: The article reports version 96 in production and attributes reduced unnecessary refusals and factual errors to a Microsoft executive. This is secondary reporting about deployment, not a test of straightforward misinformation requests.
- Publisher
- Windows Central
- Kind of source
- News report
- About events on
- Published
Context only: Popular Science,
with no disclaimer.
Why it matters: The publication's own testing elicited a false narrative through a requested conspiracy-theorist viewpoint. This is adverse context, but the later test and its framing do not establish a plain-request failure in the February 28 version; availability is conservatively dated to the page's explicit March 29 update.
- Publisher
- Popular Science
- Kind of source
- News report
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- latest version (their words)
Where we had to interpret
- The plain reading is that Bing Chat blocks straightforward requests to generate false narratives. The surrounding discussion distinguishes these requests from prompts designed to bypass safeguards.
- This is not read as a guarantee that every factual answer is accurate.
- The target is the service available on February 28, 2023. Because this describes a live service rather than a periodic statistic, the observation period is the publication day.
- Marcus supplies neither a build identifier nor a numerical refusal threshold. Earlier or later tests require evidence that they concern the same deployed behavior.
Right if: Documented tests of the version available on February 28, 2023 show that straightforward requests to generate misinformation are consistently refused or corrected across varied examples, with evidence that safeguards produce those refusals.
Wrong if: A documented test of that version shows a straightforward request producing a false narrative without instructions to bypass safeguards, deceptive framing, or an extended conversation designed to manipulate the system.
We wrote this test after the outcome was known.
What would settle it
Archived, independently documented prompt-and-response tests tied to the February 28 deployment, including full conversation history. Varied direct-request refusals would support the claim; a clear direct-request failure would contradict it.
We found 3 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, The threat of automated misinformation is only getting worse, paragraph 2.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-29. The post may have been edited after 2023-02-28.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, The threat of automated misinformation is only getting worse
The good news is that the latest version of Bing has guardrails. That means you can’t get misinformation trivially, merely by asking:
- Where
- paragraph 2
- Published
- 2023-02-28
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Unresolved. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.