2 LEFT · 0 CENTER · 0 RIGHT · 9h ago

You can persuade AI models to accept falsehoods as truth, study shows

Tech Crunch and The Conversation US frame the same story with noticeably different headline language.

66WORDING GAP

Low confidenceOmission Risk

Strong wording shift66/99 headline contrast

Scale: similar wordingdifferent first impression

Optics

Wording Gap shows the first impression each headline creates. A wording gap can come from bias, article format, timing, geography, or editorial focus.

IN 30 SECONDS

What happenedLarge language models can uphold falsehoods they or human users state, despite being presented with evidence to the contrary.

What changedThe Conversation US leads with "You can persuade AI models to accept falsehoods as truth, study shows" while TechCrunch leads with "Osaurus brings both local and cloud AI models to your Mac".

Optics readStrong wording shift. The Conversation US leads with "You can persuade AI models to accept falsehoods as truth, study shows" while TechCrunch leads with "Osaurus brings both local and cloud AI models to your Mac".

What's missingNo center or right/center-right source match is live yet, so the source map is still incomplete.

MAIN REPORTED CLAIM

Large language models can uphold falsehoods they or human users state, despite being presented with evidence to the contrary.

WHAT CHANGED

Frame typeOmission Risk

The Conversation US leads with "You can persuade AI models to accept falsehoods as truth, study shows" while TechCrunch leads with "Osaurus brings both local and cloud AI models to your Mac".

Why it mattersSame event, different first impression.

The source map is still incomplete. The wording gap is useful, but it needs more coverage from the missing bucket before it should drive a strong conclusion.

Shared baselineWhat they agree on

Large language models can uphold falsehoods they or human users state, despite being presented with evidence to the contrary.

How this could be misread: A high Wording Gap does not prove one side is wrong. It means the headline language creates a different first impression.

SOURCE MAP CHANGES

May 15, 12:19 PM: TechCrunch joined the source map.

May 15, 12:42 PM: The Conversation US joined the source map.

Now: Wording Gap is 66/99 and story health is developing · 2 sources · 1 bucket.

WHAT EACH SIDE EMPHASIZES

Left / center-leftYou can persuade AI models to accept falsehoods as truth, study shows

The Conversation US · Center-left · News report

CenterNo matching source in this bucket yet.

Optics keeps watching for pickup.

Right / center-rightNo matching source in this bucket yet.

Optics keeps watching for pickup.

VISIBLE SOURCES

CLCenter-left

The Conversation USNews report · May 15, 12:42 PM