Every community
needs a good ally.

Meet Bluey. A friendly companion with a serious purpose: helping your moderation team recognize possible criminal recruitment and give it the attention it deserves.

Watch the film1 minute, with sound

Bluey: Good nose. Good company.

Read the transcript

Every community needs a good ally.

Meet Bluey: a friendly Discord companion with a serious job, helping your moderators spot possible criminal recruitment.

It rarely looks like crime. It looks like an opportunity.

Bluey reads the channels you choose and notices the clues.

An offer of easy money. A request for someone young. A push to go private.

Even when the words are disguised.

Then it quietly flags the message to your team, with its reasons attached.

Your team makes the call. A flag is a reason to look closer, never a verdict.

For direct messages, there's Ask Bluey, now in private testing. Pick one message; the answer is just for you.

Bluey. Good nose. Good company.

Meet your new ally at getbluey.app

An illustrative film with fictional messages. Ask Bluey is in private testing and cannot be installed yet.

A little extra support for the people who care.

Good nose.
Good company.
Your community
moderationPrivate channel
BlueyAPP

This message could use a closer look.

generalnova · 14:02
Easy money for a delivery tonight. Need someone young. Message me privately.
  • Payment offer
  • Youth targeting
  • Private contact
SamModerator

I’ll check the context. Thanks, Bluey.

Illustrative Discord conversation.

An ordinary message.
A reason to look closer.

The offer. The pressure. The request for privacy. See how separate clues become a review your team can explain.

It starts like an opportunity. The details change the picture.

A payment offer alone tells you little. Add youth targeting and a request for private contact, and your team has a clearer reason to review.

general
nova
Easy money for a quick delivery. Need someone young who can go tonight. Message me privately. Keep this between us.

Read the whole message. The urgency and request for secrecy also matter to the human review.

Bluey’s assessment

A combination worth investigating

Bluey surfaces the payment offer, youth targeting and private contact together. A moderator can inspect those reasons and check what the “delivery” refers to.

  • Payment offer
  • Youth targeting
  • Private contact

The flag includes its reasons. Your team decides whether to investigate, ask for context or dismiss it.

Illustrative examples, not a live scan. A flag is a reason to investigate, not a finding of wrongdoing.

A scientific approach.
Reasons you can inspect.

Heuristics are explicit rules for recognising patterns. Bluey brings those clues together into an assessment your team can question.

One connected team. A shared understanding.

Bluey brings content from selected Discord channels to Blue. Blue analyses the language and returns signals and context. Bluey shares findings with moderators, who review them and record their decision and feedback alongside the original content and assessment.

Your communitySelected Discord channels

Bluey

Your companion

Blue

The analysis engine

ContentSignals & contextFindings for reviewDecision & feedback
A shared audit trail

Original content, analysis and human review. Connected in one record.

Rules can miss recruitment or flag innocent messages. A signal starts a review; it doesn’t establish intent.

Someone you can
always ask.

A moderation team only sees what reaches a channel. Sometimes the approach that matters most arrives as a direct message.

DMs are outside of the area where Bluey can monitor and protect, but you can always: Ask Bluey.

kaiden_ttv

This is the beginning of your direct message history with @kaiden_ttv.

Yesterday

kaiden_ttv

hey, saw you in the server earlier

you’re on most nights right?

Today

kaiden_ttv

Easy money for a quick delivery tonight. You seem like someone I can trust. Don’t tell anyone, ok?

You used Ask Bluey

BlueyAPP

This message looks a lot like the way people recruit for crime. Take it seriously.

What Bluey noticed:

  • It offers money for something vague.
  • It pushes you to act right now.
  • It flatters you before asking for something.
  • It asks you to keep it secret.

What you can do:

  • Do not agree to anything and do not go anywhere to meet them.
  • Screenshot the chat before you block, so you have it if you need it.
  • Tell an adult you trust today. If you would rather talk to someone outside your family, BRIS is at 116 111.
  • If someone is in danger right now, call 112.

Bluey only saw what you picked, nothing before or after it.

This check was private. Nobody else was told, and Bluey did not keep the message.

Only you can see this · Dismiss message

  1. The moment it feels off.

    The approach arrives in a direct message, late, to one person. There is no channel to report it in and nobody obvious to ask.

  2. A right-click, nothing more.

    Bluey is not sitting in the conversation waiting. It is opened from the message itself, once, by the person reading it.

  3. One message, handed over.

    Under Apps, Ask Bluey receives exactly the message that was picked. Not the conversation, not the inbox, not the account.

  4. An answer meant for them.

    What Bluey noticed, what they can do next, and what it could not see. Nobody else is told and the message is not kept.

Ask Bluey

A friend that is always available

Independent from the server bot, Ask Bluey is meant for individual use, it provides a surface area for them to get help from Bluey to navigate difficult questions and situations.

  • Ask about a messageAsk Bluey doesn’t read anything in the background, the ask is invoked by the user
  • The answer goes directly to themIt needs to be psychologically safe to ask, no reports are sent to any moderator or parent
  • No messages are keptThe message that they ask about is not stored, it’s a direct support from Bluey, not a follow-up chain

Ask Bluey is in private alpha, if you want to be involved, reach out!

I want to know more

Illustrative example with fictional content, not a live check. Bluey can miss an approach or draw attention to a harmless message.

Better learning starts
with better expertise.

The people closest to recruitment understand what a phrase means, how approaches change, and when a warning looks like a threat.

We’re developing a learning model around that insight: expert-authored scenarios, independent review and evaluation on examples the model hasn’t seen.

The learning model is in development. Institutional contributors are welcome.

Bring your expertise
Our model-development approach
  1. 1

    Professional insight

    Safeguarding, crime-prevention and research professionals contribute realistic, authored scenarios.

  2. 2

    Independent review

    A second expert reviews the interpretation. Uncertain examples stay out of training.

  3. 3

    Learning and evaluation

    Candidate models learn from approved examples, then face unseen cases, benign messages and disguised language.

  4. 4

    A considered release

    Compare missed signals and false flags. Review the evidence before approving a new model version.

Human-reviewed examples. Traceable versions. Learning with oversight.

“It found something”
is only half the test.

A useful model must also recognise when a message is harmless. Our evaluation plan covers missed recruitment, false flags, disguised wording and legitimate safety discussions. Training and test examples are separated by scenario family to make the comparison more meaningful.

The research behind the questions

Text-perturbation research motivates testing beyond clean, predictable wording. NIST’s AI Risk Management Framework informs our emphasis on evaluation, domain expertise and human oversight.

A few good questions.

Who is Bluey for?

Bluey is for Discord community owners and moderation teams who want help recognising possible criminal recruitment. Get in touch to discuss your community and current availability.

Does Bluey replace our moderators?

Your team remains in charge. Bluey surfaces possible recruitment and the signals behind its assessment. People review the context and decide what to do. Bluey can miss content or flag something harmless.

Can Bluey read private conversations?

The server bot assesses content in the server channels selected and accessible to it. It does not provide passive access to members’ private direct messages. Ask Bluey works the other way round: the person decides which single message it sees.

What about a message nobody else sees?

That is what Ask Bluey is for. A person installs it to their own Discord account and shows it one message at a time; the answer goes to them. There is no monitoring, no guardian view and nothing stored. Ask Bluey is in private testing and cannot be installed yet.

How do we get started?

Tell us about your community at hello@getbluey.app. We can walk you through Bluey and discuss the current setup and availability.

Your expertise.
A safer community.

Work in safeguarding, research, crime prevention or community moderation? Let’s explore what Bluey could learn from your perspective.

Share your expertise