Athena: Next Generation Known and Unknown CSAM Detection
First-generation CSAM is evolving rapidly. Athena delivers the real-time detection and classification needed to respond to today’s threat landscape and set a proactive, defensible standard of care.
Detect known and unknown CSAM in real time
Athena detects known and previously unseen CSAM in real time, including synthetic and AI-generated material. Integrated directly into platform workflows, it enables immediate action so organizations can identify threats earlier, respond faster, and protect users at scale.
The end of the ‘known’ era
The IWF’s reported 26,385% surge in AI-generated abuse material marks a turning point for digital safety. With no hashes to match new content, we must move beyond legacy detection and close the visibility gap.
Harm Proliferation
CSAM is growing at an unprecedented rate, driven by a surge in hybrid threats, including generative AI, coerced “self-generated” content, and global financially-motivated extortion networks.
The Detection Blindspot
Traditional hash-matching tools are unable to detect altered or first-generation synthetic content, leaving platforms exposed to novel harms.
Human Cost
Predators weaponize first-generation content to mask their digital footprints, slipping through the blind spots of conventional safety tools.
Two types of CSAM. Two detection approaches.
Known CSAM
Material previously identified, verified, and classified by safety organizations such as NCMEC or law enforcement.
- Hash Matching: Digital fingerprints of uploaded content are compared against established databases in milliseconds. When a file matches, it is flagged immediately.
- The Limit: Hash matching is reactive by design. It identifies content that has already been seen — and only that content. Any file that has not been previously catalogued produces no match and no alert.
Unknown CSAM
Newly created, modified or AI-generated content that has not been previously seen or catalogued — and therefore has no hash.
- AI Detection: Athena analyzes deep visual signals and structural patterns in real time, identifying the characteristics of abuse in content that has never been encountered before. It does not search for a known file. It assesses what is in front of it.
- The Value: This is what moves detection from reactive to proactive. Athena identifies content at the point it appears on a platform — before it can be catalogued, redistributed or cause further harm.
The next generation detection classifier
Powered by technology & people
How Athena detects the ‘undetectable’
Athena, powered by the Roke Vigil AI CAID Classifier, moves beyond traditional hash matching by combining the speed of digital fingerprinting with the predictive power of law enforcement-grade AI. By analyzing deep visual patterns rather than just file metadata, Athena identifies harmful content at the point of creation—detecting “unknown” threats before they have ever been seen, hashed, or cataloged by a database.
Future-proof harm management
A multi-layered defense that bridges the gap between today’s operations and tomorrow’s legal mandates.
CAID-Powered Intelligence
Resolver leverages the Roke Vigil AI classifier to identify synthetic and novel content that standard hash matching misses. Trained on the UK’s Child Abuse Imagery Database (CAID), our model provides unmatched forensic accuracy and triple-check law enforcement verification to ensure your detections are evidence-grade.
Graded Detection
Athena is the only model that distinguishes content by severity (Categories A, B, and C). By combining low-latency processing with severity-based detection, we enable prioritized triage and immediate content takedowns, stopping the spread of high-harm abuse in milliseconds.
Adaptive Retraining
Athena is regularly retrained to identify evolving abuse tactics, AI-generated material, and coerced “self-generated” content. By pinpointing grooming-related media and “age-play” triggers, Athena ensures predatory behaviors cannot hide in plain sight, disrupting abuse at the source.
Future-Proof Compliance
Anticipate the shift in global mandates. While hash-matching is the current baseline, Athena prepares you for a future where proactive detection of “Unknown” CSAM is expected. Document your mitigation efforts with precision, transforming compliance from a legal burden into a competitive advantage.
Duty of Care
Eliminate the psychological burden on your internal staff by outsourcing sensitive content review to Resolver’s dedicated specialists. Our Graded Classification ensures that only the most necessary cases are escalated, while our expert-led review team handles the trauma-exposed work, allowing your internal T&S team to focus on high-level strategy and platform integrity.
Secure & Flexible Deployment
Deploy via secure API or on-premises models to fit your infrastructure. Our zero-image retention policy ensures your data is never used for retraining; Athena is a closed-loop system powered only by CAID-verified material and licensed negative adult data.
"When your platform is flooded with novel CSAM content, your safety team's time is stolen from the children who need rescue most. Athena’s graded detection doesn't just 'find' images - it triages them, separating synthetic noise from high-priority forensic evidence so investigators can focus where it matters: saving real lives."
Director of Product, Trust & Safety, Resolver
Athena's multi-layered defence pipeline
Known Detection: Perceptual Hashing
High-performance matching engine for your platform’s existing hashing workflows. By utilizing industry-standard fingerprints, Athena performs sub-millisecond comparisons against verified global databases.
Unknown Detection: AI Classification
When no hash match match exists, Athena provides superior detection by leveraging the unparalleled quality of training data and direct training access to the world’s most highly curated imagery dataset (CAID) in a secure laboratory environment. The classifier is routinely re-trained to ensure that the model continues to reflect evolving abuse dynamics.
Specialist Human Review
To ensure absolute reporting accuracy, Athena can route high-confidence flags to Resolver’s specialist human reviewers, providing a final layer of verification.
Get started with Athena
Run a proof-of-concept to verify Athena’s graded detection and forensic accuracy in identifying unknown CSAM.