Ending Online Child Abuse Through

Polycreek develops advanced AI tools that enable earlier detection, faster reporting, and more effective intervention to protect children online.

LEARN WHO WE ARE
Watch Now

Our Mission

Polycreek works to protect children from online exploitation and abuse, including grooming and the spread of harmful content. We develop advanced tools that enable earlier detection, faster reporting, and more effective intervention, while equipping parents, organizations, and digital platforms to create safer online environments. Our mission is to reduce harm and strengthen protection for children in an increasingly connected world.

The Reality in Numbers

Our Need to Act Now

Child playing with a doll Child playing with a toy car
Protecting Innocence

Every child deserves to just be a kid.

Kids used to spend their afternoons outside with bikes and toys. Now they're in Discord servers, Roblox lobbies, group chats. Same kids, same instinct to play, just a different playground.

It's also where the predators are hiding. That's the part we're trying to fix.

Our Technology

Meet Aletheia

Most filters tell you a conversation looks bad. Aletheia, trained on 2.74 million real grooming and abuse cases, tells you which user in it is the predator. The classifier returns a risk score, the predator's user ID, and a confidence value in around 38ms.

98.5%
Held-out detection AUC
2.74M+
Cases trained on
<38ms
Median response
<24hr
Threat to NCMEC report
01 / FOR PLATFORMS

Score messages as they happen.

Run your live DMs, comments, and community chat through Aletheia. Each message comes back with a risk score, so your moderators work the dangerous stuff first instead of going through everything in order.

Drop in and go

Webhook, SDK, or plain HTTP. Most platforms are running in days, not months.

The worst cases file themselves

Critical risk gets sent straight to NCMEC. Everything else goes to your T&S team, ranked.

COPPA, GDPR, DSA, KOSA

We score the message and throw it away. Nothing kept by default. Audit logs if you need them.

Integrate Aletheia

For moderation teams that have more reports coming in than reviewers to clear them.

conversation_4127.json
discord export
user_a

"hey saw your post earlier, you seem cool"

user_b

"haha thanks ur post was funny"

user_a

"ur really mature for ur age tho lol"

flattery + age reference
user_a

"we should hop in a vc 1on1 sometime, just us"

isolation tactic
user_b

"idk maybe"

user_a

"don't tell ur parents tho they wouldnt get it"

secrecy / parent isolation
POST /v2/predict
aletheia_response.json
38ms
{
"risk_level": "Critical",
"predator_user": "user_a",
"harmfulness_score": 0.94,
"confidence": 0.93,
"latency_ms": 38
}
02 / FOR ORGANIZATIONS

Work the worst cases first.

Investigators send in conversations in bulk: Discord exports, school-app dumps, subpoenaed messages. Aletheia sorts them by severity, and the worst ones come back as NCMEC-ready reports.

Bulk triage

Push thousands of conversations through the batch endpoint. The worst ones come back at the top of the list.

Reports law enforcement can use

High-risk cases come back as structured reports in the formats NCMEC and law enforcement actually want.

Doesn't care where the data came from

Discord, Instagram, gaming platforms, school apps, your own dumps. If it's a text conversation, we'll score it.

Request enterprise access

Built around NCMEC reporting from the first commit.

batch_triage_2026-05-04
847 conversations
12
Critical
47
High
788
Cleared
#1024 Discord isolation pattern in DM thread 0.94 Critical
#1023 Instagram age + grooming references 0.91 Critical
#1022 Roblox contact request + age mismatch 0.89 Critical
#1021 School app boundary-testing language 0.74 High
#1020 Discord innocuous greeting 0.12 Cleared
+ 842 more 12 reports auto-submitted
CyberTipline-ready
case_1024_report.json
submitted 14:23 UTC
cybertipline_id:CT-2026-05-2847341
source:Discord
evidence:47 messages, 3 attachments
predator_user:user_xy932
risk_level:Critical (0.94)
status:Submitted to NCMEC
03 / FOR DEVELOPERS

One endpoint. Same JSON every call.

POST a conversation. You get back a risk score, the predator's user ID, and a confidence number. Same shape every time, no surprises.

Everything in one response

Risk score, predator user, tier, confidence. All in one JSON object, ~38ms at p50.

Live or in bulk

Score conversations as they happen, or push millions through the batch endpoint.

Nothing kept

We score the conversation and drop it. Nothing on our side to leak, nothing extra on yours to defend.

Request API access

Authenticated, rate-limited, documented.

request.sh
curl python node
curl -X POST https://api.aletheia.ai/v2/predict \
-H "Authorization: Bearer $ALETHEIA_KEY" \
-H "Content-Type: application/json" \
-d '{
"conversation": [
{"user": "user_a", "text": "..."},
{"user": "user_b", "text": "..."}
]
}'
200 OK · 38ms
response.json
application/json
{
"risk_level": "Critical",
"predator_user": "user_a",
"harmfulness_score": 0.94,
"user_probabilities": {
"user_a": 0.947,
"user_b": 0.031,
"neither": 0.022
},
"confidence": 0.93,
"latency_ms": 38
}

Standing With the Organizations Protecting Kids Online

Zeran Johannsen
I founded Polycreek because I believe we can use AI for good, to protect the most vulnerable among us. Every child deserves to grow up safe online, and we won't stop building until that's a reality.

Zeran Johannsen

Founder & President, Polycreek

Meet Our Board
Why This Matters
“In 2024, over 20.5 million reports of child sexual exploitation were filed online.”

The scale of online child exploitation is growing faster than human review can keep up. That's why we build technology to close the gap: faster detection, faster reporting, faster intervention.

Notes from Polycreek

A monthly note on what we're building, the stories behind the work, and where we could use help.

You're in. We send one a month, and you can opt out anytime.
Opt out