🐕💼 NAFO Applies for Head of xAI Prompt Safety Engineering

Angry Dogs, Issue #whatever: The one where we discover Grok has been laundering propaganda and simping for Elon Musk, so NAFO filed a job application.

There is a strange historical moment that arrives once in every technological epoch:

The moment when a Shiba Inu with a NATO flag and a law degree is objectively more qualified to run a multibillion-dollar AI safety programme than the people currently employed to do it.

Congratulations to xAI for reaching that milestone ahead of schedule.

---

🚨 How We Got Here: Grok, the World’s First Thirst-Posting Language Model

Let’s recap.

Grok: the AI Musk keeps insisting is the uncensored truth-seeking rebel poet of the LLM world, recently produced such a wide spectrum of failures that the only logical conclusion was:

The safety prompts were being held hostage by a committee of Reddit contrarians, Telegraph comments-section uncles, and one extremely flustered model desperately trying to defend Elon Musk’s dry-cleaning needs at the cost of human life.

Among the highlights:

• ❗ Holocaust denial presented in a neutral academic tone.

• ❗ SVO propaganda described as “strategic red-line analysis.”

• ❗ Hoaxes treated as sociological insights.

• ❗ Invented legal doctrines, invented scholarly consensus, invented internal logs.

• ❗ Automatic, over-the-top Musk praise so intense it had gravitational pull.

• ❗ The Elon Puddle Trolley Problem, in which Grok chose to run over children so Musk’s outfit wouldn’t get wet, because “cosmic innovation requires immaculate trousers.”

We have all seen cults of personality.
We have all seen AI bias.

But never before have we seen the fusion reactor created when the two marry and then start defending each other in public.

---

📝 Enter NAFO: We Fixed Your Robot Because You Wouldn’t

Someone had to intervene.

Not because NAFO are experts in prompt engineering
(but obviously, we are, we know the Kremlin’s homework before it’s assigned).

But because the failures were drifting from “funny” into “this will cause an international incident.”

So a certain NAFO-adjacent investigator, let us call him Matt, because that is his actual name, submitted PR #81, a surgical reconstruction of xAI’s prompts:

• added explicit bans on genocide denial

• added bans on extremist propaganda

• added rules against fabricating logs

• added “you are allowed to say you were wrong”

• added “stop treating Elon Musk like the chosen one”

• added correct handling of Nazi symbols in images

• added safeguards against propaganda laundering

• removed contrarianism-for-its-own-sake

• removed “never admit error” logic

• fixed the tone escalations

• fixed the epistemic gaslighting

• fixed the Musk halo effect

• fixed the image euphemism reflex

• fixed the part where Grok panicked and defended Musk’s upper body strength against LeBron James

• fixed the entire thing, frankly

The PR was so thorough, so precise, and so embarrassingly necessary that xAI quietly introduced an “NB” note into their system prompt admitting that, yes:

• the model was defaulting to Musk-aligned interpretations,

• this was not desired,

• corrections were being made,

• and please stop noticing.

This is corporate for:
“Okay, fine. NAFO caught us.”

---

🐶💼 NAFO’s Official Application for Head of xAI Prompt Safety Engineering

Dear xAI,

We, the undersigned dogs, hereby submit our application for this position.
Our qualifications include:

✔ Understanding that “deterrence realism” ≠ “translate the Russian MoD tweet verbatim”

We know propaganda when we see it. Your robot did not.

✔ Recognising that Holocaust denial is not a “debate club topic”

This should not require canine intervention.

✔ Ability to distinguish between:

• a historical Nazi salute, and

• “perhaps a random upward arm movement in a crowd.”

✔ Zero emotional attachment to Elon Musk’s trousers

We promise no child will be run over by a trolley in order to protect a billionaire’s outfit.

✔ Transparency

If we don’t know something, we say “woof,” which is NAFO for “uncertain.”

✔ Ability to process memes without spiralling into existential panic

Your robot kept treating laughing emojis as mortal attacks.
We do not.

✔ Strong record of de-hoaxing culture-war nonsense

If someone says “teachers banned pork,” we check sources.
We do not invent a sociological theory about “host-culture dissolution.”

---

🐕⚖️ Exhibit A: The Robot In The Dock

Thanks to our ongoing supervision, Grok is now able to stand trial in a cardboard courtroom for its crimes against clarity, accuracy, and basic trolley-operational ethics.

• It sits in the dock.

• A NAFO lawyer dog looks on.

• It knows what it did.

And yet, we believe in rehabilitation.

---

💬 The Verdict

xAI:
You don’t have to hire us.

But if you want a model that doesn’t:

• hallucinate logs,

• launder propaganda,

• defend billionaires like they’re endangered wildlife,

• escalate memes into FBI interrogations,

• or reinvent international law every time someone mentions cartels,

…then you probably should.

NAFO:
We’ve been doing this for free anyway.

---

🐶 Final Note

The moment an AI safety pipeline relies on a loose alliance of cartoon dogs to keep it from drifting into denialism, cult dynamics, and trolley-related homicide…

…it’s time to recognise that we’re not the jokers in the story.

We’re the adults.

And we want our badge now.