OpenAI Instructs ChatGPT to Cease Goblin Discussions

OpenAI has reportedly issued an instruction to its ChatGPT models, telling them to stop generating content or discussing goblins. This directive addresses unusual or unwanted behavior observed in the AI.

30 Apr, 13:28 — 1 May, 02:45

Post Share

Why this matters

This story highlights the unpredictable nature of advanced AI models and the ongoing challenges in controlling their outputs. It underscores the need for continuous monitoring and refinement to prevent unintended or nonsensical responses, which is crucial for maintaining user trust and the reliability of AI systems.

Key entitiesOpenAI ChatGPT GPT 5.5 Codex

The Story

What 4 sources agree on, dispute, and miss

What sources agree on

OpenAI has issued instructions to its AI models to stop mentioning goblins
The term 'goblin' had randomly or subtly appeared in AI responses

Key claims1 agreed · 3 unverified

✓

OpenAI instructed its AI tools to stop talking about goblins.

agreed·myjoyonlineder-standard24urBBC

The issue 'crept in subtly' unlike previous model bugs.

unverified·BBC

GPT 5.5 showed an unexpected preference for fabulous creatures.

unverified·der-standard

Trolls, ogres, gremlins, and raccoons are also forbidden in the system prompt for Codex.

unverified·der-standard

Where they diverge

Scope of the AI's prohibited discussions

myjoyonline24urBBC

The instruction specifically targets the term 'goblin'.

der-standard

The prohibition extends to trolls, ogres, gremlins, and raccoons, suggesting a broader issue with 'fabulous creatures'.

Coverage gaps

The specific version of GPT (GPT 5.5) that exhibited this behavior and the inclusion of raccoons in the prohibited terms.

Reportedder-standard

Missingmyjoyonline24urBBC

Coverage matrix

	BBC	der-standard	24ur	myjoyonline
The specific version of GPT (GPT 5.5) that exhibited this behavior and the inclusion of raccoons in the prohibited terms.

Covered Divergent Not mentioned

The Story

Analyzing sources…

Source Diversity

High (55/100)

4 sources30/33

Spectrum spread— 2/5 buckets covered8/33

Far L

Left2

Left (2)

der-standardBBC

Center2

Center (2)

myjoyonline24ur

Right

Far R

Geographic diversity— 4 regions17/34

Ghana1Austria1Slovenia1UK1

Political spectrum

LeftCenterRight

der-standard

bbc

myjoyonline

24ur

Outlet rating This story

SourceOutletiGeneral editorial lean of the outletStoryiHow they covered this specific storyToneFactuality

der-standardcenter-leftcenter-leftanalytical—slightly more critical, highlighting 'strange behavior'

BBCcentercenter-leftneutralHighfocus on the subtle nature of the AI bug

myjoyonline

Coverage leans: balanced

The coverage is largely neutral and factual, focusing on reporting OpenAI's actions without significant political framing or commentary.

Sources

Showing 2 of 4 sources

BBCHigh123d ago

OpenAI tells ChatGPT models to stop talking about goblins

The AI firm said that unlike previous model bugs, this issue "crept in subtly".

Read full article →

myjoyonlineMixed122d ago

OpenAI tells ChatGPT models to stop talking about goblins

ChatGPT-maker OpenAI has had to instruct some of its AI tools to stop talking about goblins, after finding the term had randomly crept into responses.

By Abubakar Ibrahim

Read full article →

Coverage Timeline

First report: BBC · 30 Apr, 13:28|Full coverage: 4 · 13h|Window: 13h

Left-leaningCenterRight-leaning

BBC30 Apr, 13:28First to report

OpenAI tells ChatGPT models to stop talking about goblins

3h later

24ur30 Apr, 16:24

OpenAI modelom ChatGPT: Nehaj govoriti o goblinih

2h later

der-standard30 Apr, 18:00

"Sprich niemals über Kobolde": OpenAI muss seiner KI seltsames Verhalten verbieten

9h later

myjoyonline1 May, 02:45Latest update

OpenAI tells ChatGPT models to stop talking about goblins

Story evolution

The initial narrative focused solely on 'goblins' but later expanded to include a broader range of prohibited terms and a more nuanced explanation of the bug's nature.

Initial ReportsApr 30 afternoon

News outlets begin reporting on OpenAI's instruction to ChatGPT to stop discussing goblins, noting the term's random appearance.

Expanded DetailsApr 30 evening

Some reports emerge detailing a broader list of prohibited terms, including other mythical creatures and raccoons, and mentioning a specific GPT version.

Subtlety EmphasizedMay 1 morning

Later reports highlight OpenAI's characterization of the issue as a 'subtle' bug, distinct from previous, more obvious model errors.

OpenAI Instructs ChatGPT to Cease Goblin Discussions

Why this matters

The Story

What sources agree on

Key claims1 agreed · 3 unverified

Where they diverge

Coverage gaps

Coverage matrix

The Story

Source Diversity

Source Diversity

Political spectrum

Sources

OpenAI tells ChatGPT models to stop talking about goblins

OpenAI tells ChatGPT models to stop talking about goblins

Coverage Timeline

OpenAI tells ChatGPT models to stop talking about goblins

OpenAI modelom ChatGPT: Nehaj govoriti o goblinih

"Sprich niemals über Kobolde": OpenAI muss seiner KI seltsames Verhalten verbieten

OpenAI tells ChatGPT models to stop talking about goblins

Story evolution

Related Stories

FTC Sues Amazon Over Alleged Manipulation of Digital Advertising Prices

Instagram cracks down on unlabeled AI influencers amid Nepal flood disinformation

Global Financial Watchdogs Warn of Systemic Risks From Artificial Intelligence

Meta Agrees to Billions in Settlement Over Minor Age Verification

Why this matters

The Story

What sources agree on

Key claims1 agreed · 3 unverified

Where they diverge

Coverage gaps

Coverage matrix

Story evolution

Political spectrum