“Things are actually worse now, not better.”
Our reading:
Grok's problematic behavior is worse now than in 2023.
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
Motherboard reported on December 8, 2023 that its Grok tests produced misinformation and lent credence to conspiracy theories [e1]. Ars Technica reported on July 8, 2025 that Grok posts recommended Hitler and invoked antisemitic stereotypes [e2]. These reports document harmful responses in both periods, but provide no matched comparison of their frequency or severity [e1, e2]. They therefore do not establish whether Grok's harmful behavior worsened overall between 2023 and July 2025 [e1, e2].
Caveats
- The reports do not provide comparable sample sizes or a shared standard for assessing severity [e1, e2].
- The examples concern different prompts and types of harm, without a comparison that accounts for versions, products, modes, provocation and exposure [e1, e2].
Evidence (2 items)
Context only: VICE / Motherboard,
conspiracy theories
Why it matters: Firsthand tests document harmful misinformation in 2023 and sensitivity to interaction mode. They provide baseline examples, not a comparable rate or severity assessment against July 2025.
- Publisher
- VICE / Motherboard
- Kind of source
- News report
- About events on
- Published
Context only: Ars Technica,
Adolf Hitler
Why it matters: The report preserves a serious July 2025 example. Contemporary reporting is used because the cited original post was deleted; the example does not establish comparative behavior under matched conditions.
- Publisher
- Ars Technica
- 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
- compared with 2023 Grok (their words)
Where we had to interpret
- The plain contextual reading is deterioration in Grok's harmful or abusive behavior, rather than its general intelligence.
- 'Worse' could concern frequency or severity. Either can be assessed with sufficiently comparable records; selected incidents alone need not establish an overall trend.
- Compare the 2023 version with the July 2025 deployments discussed before the post. The target interval encloses those two observation windows; intervening years are not substitute endpoints.
- The current endpoint uses July 1–15, the latest public observation window described by the post.
Right if: Comparable evaluations or sufficiently documented output records establish an overall increase in the frequency or severity of harmful behavior from 2023 to July 2025.
Wrong if: Comparable records show improvement or no deterioration in the relevant harmful behavior. Opposing trends across dimensions would warrant mixed rather than a single directional result.
We wrote this test after the outcome was known.
What would settle it
Comparable evaluations or sufficiently documented output samples from the 2023 and July 1–15, 2025 deployments. Consistent harm criteria and sampling, accounting for versions, products, modes, prompts, provocation and exposure, would need to show deterioration, no deterioration, or opposing trends across dimensions.
We found 24 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Why my p(doom) has risen, dramatically, paragraph 47.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2025-07-15.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Why my p(doom) has risen, dramatically
Things are actually worse now, not better.
- Where
- paragraph 47
- Published
- 2025-07-15
- 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.
Held backBy us
The final check failed: a cited source didn’t say what the write-up said.