Protecting online conversations with AI, putting child safety at the forefront. CQcjASdDAvTMJsybZTJJ3hvfdwbi4uTqmtd2SNqYpump $SC
Joined January 2025
- Tweets 364
- Following 7
- Followers 1,239
- Likes 178
85 Photos and videos
Pinned Tweet
19 Sep 2025
SafeCircle: The Invisible Protection Every Family Needs
2
6
40
9,944
Jun 12
Horizon v2 is here.
Built for child safety. Designed for privacy.
huggingface.co/safecircleai/β¦
2
7
10
125
May 27
Big things are coming π
Our SafeCircle web dashboard is taking shape giving parents real-time insights, alert management, and peace of mind in one sleek interface.
This is just the beginning.
4
2
9
565
May 20
Proud to be featured by Silicon Snark for our privacy-first approach to child safety. SafeCircle is building protection without turning privacy into collateral damage: siliconsnark.com/safecircle-β¦
6
9
16
399
May 19
Horizon is now available to the public to test. Take a look at the license before trying our models safecircle.tech/en/licenses/β¦
All our models are available at huggingface.co/safecircleai
1
2
8
301
SafeCircle retweeted
May 18
The new version of horizon will include a much bigger dataset with 1.6 million conversations to reduce false positive and improve the performance of our in house model. All while protecting privacy and not training on user data.
1
1
172
May 12
Announcing Horizon β our on-device AI that detects grooming, bullying & threats in real-time on your child's phone.
25MB model. Sub-100ms inference. Zero messages sent to any server.
Privacy isn't our policy. It's our architecture.
Read the full paper: safecircle.tech/posts/releasβ¦
#ChildSafety #AI #Privacy
5
9
19
437
May 12
Cyberbullying doesn't always look like obvious threats.
It's exclusion. Rumours. Screenshots shared without consent. Group chats where one person is the target.
SafeCircle's AI detects patterns β not just keywords.
#CyberBullying #ChildSafety
1
3
165
May 12
85% of parents say they're worried about their child's online safety.
But most feel they don't have the right tools.
We built SafeCircle to change that.
π‘οΈ safecircle.tech | #OnlineSafety #AI
1
5
163
May 11
Schools in Europe: did you know you can deploy SafeCircle across your entire student body for β¬8β20 per student per year?
GDPR-compliant. Centralized dashboard. Real-time alerts to safeguarding staff.
Book a demo β safecircle.tech
#SchoolSafety #EdTech
3
114
May 11
At SafeCircle we obsess over one question:
How do we detect real threats without generating so many alerts that parents and teachers tune them out?
It's the hardest calibration in the product. And the most important.
3
118
May 11
Keyword filters: block "meet me tonight"
AI conversation analysis: understands when "let's keep this between us" from a stranger is a red flag
That context is the difference between catching a threat and missing it.
That's SafeCircle.
1
1
5
194
May 10
The question every school administrator should be asking:
"If a predator was grooming one of our students right now β would we know?"
Most wouldn't. Not until it was too late.
SafeCircle changes that.
1
1
108
May 10
Schools have a legal duty of care for students online.
Most have no tools to fulfil it.
6/10 victims never self-report.
Content filters miss conversation-based threats.
Manual review doesn't scale.
SafeCircle fills the gap. Automatically.
π‘οΈ safecircle.tech/schools | #SchoolSafety #EdTech
1
77
May 9
Privacy shouldn't be a trade-off in child safety.
SafeCircle's AI detects threats without storing or selling your child's conversations.
GDPR-compliant. Privacy-first. Always.
π‘οΈ safecircle.tech
1
4
124
May 9
Parents: does your child know what to do if an adult online asks them to keep a secret?
That one question could be the most important safety lesson you give them.
(SafeCircle handles the 24/7 technical side β but this conversation matters π)
2
102
May 8
6 out of 10 kids who experience cyberbullying never tell their parents.
Not because they don't trust them.
Because they don't know how.
That silence is the real danger.
#ChildSafety #CyberBullying
3
86
May 7
750,000 online predators are active every single day.
A child can be contacted in as little as 15 minutes.
SafeCircle's AI detects grooming and cyberbullying in real time β before harm happens.
Because 15 minutes is not enough time to wait.
π‘οΈ safecircle.tech
1
2
116
Apr 18
Privacy protection starts with architecture.
SafeCircle uses:
π Federated learning β AI improves without accessing your data
π End-to-end encryption β Your messages stay yours
π± Edge computing β Processing happens on your device
π‘οΈ Differential privacy β Statistical patterns, not individual data
These aren't buzzwords. They're your guarantees.
#PrivacyByDesign #SecureByDefault #TechArchitecture
1
4
106
Apr 17
How our AI works without reading your messages:
π Detects concerning behavior, not keywords
π Looks for context and relationships
β‘ All processing happens on-device
Think of it like a smoke detector:
It doesn't record everything in your house. It just alerts you when there's danger.
Smart protection. Complete privacy.
#AIForGood #PrivacyPreserving #TechExplained
2
73