CalypsoAI’s August release delivers powerful new capabilities across its AI Security Platform, enhancing observability, compliance, real-time defense, and red-teaming to keep enterprises ahead of evolving threats.

CalypsoAI AI Security Platform

Splunk Integration

We’ve introduced native Splunk integration to simplify how enterprises operationalize AI security. Instead of relying solely on API-based connections, Splunk users can now ingest CalypsoAI logs directly. This enables:

• Centralized visibility: AI threat events appear alongside your existing network, endpoint, and cloud telemetry.
• Real-time action: Trigger automated alerts and SOAR workflows the moment high-risk AI activity occurs.
• Simplified compliance: CalypsoAI logs flow seamlessly into your existing retention and audit processes.

As Splunk is a dominant SIEM in enterprise security, this direct integration reduces friction and accelerates time-to-value for AI threat detection and response.

Inference Defend

Outcome Analysis

Outcome Analysis adds unprecedented visibility into AI threat detection by showing exactly why a scanner flagged or blocked an event. Security teams can now:

• Identify the triggered scanner and the precise content or rule violation.
• Trace timestamps and user activity for audit and incident response.
• Validate scanner performance and compare behavior across versions

This feature transforms alerts into actionable insights, helping teams move from signal to resolution faster.

EU AI Act Scanner Package

CalypsoAI’s new EU AI Act Scanner Package provides immediate, out-of-the-box baseline coverage for Article 5’s “prohibited risk” categories:

1. Manipulation
2. Social Scoring
3. Biometric Data Harvesting
4. Surveillance

These scanners proactively detect and block AI behaviors that could trigger the Act’s heaviest penalties (up to €35M or 7% of global turnover). For organizations building GenAI applications in or for the EU, this package offers an essential first step toward compliance.

Custom Scanner Versioning

Security teams now have full lifecycle control of their AI scanners:

• Create unlimited versions and test them in parallel.
• Publish, roll back, or enforce versions at the project level.
• Align scanner updates with regulatory changes or emerging threats without disrupting production.

Custom Scanner Versioning turns AI security into a flexible, iterative process—keeping defenses agile and future-proof.

Inference Red-Team

Agentic Fingerprints

With Agentic Fingerprints, red teams gain full visibility into how adaptive, multi-turn AI attacks unfold. This interactive visualization shows:

• Each decision, prompt, and response during an attack.
• The Red Agent’s reasoning at every step.
• Successful, terminated, or backtracked paths across attack vectors.

This transparency allows security, compliance, and development teams to identify root causes, validate testing rigor, and prepare audit-ready reports.

Agentic Signature Attack Packs

Our monthly Signature Attack Packs are now fully agent-driven. CalypsoAI’s autonomous agent researches, generates, tests, and packages new high-impact attacks each month, ensuring:

• Continuous coverage against evolving threats.
• Faster and more scalable vulnerability discovery than human-only teams.
• Real-world validation using the same attacks we deploy against top global models.

Red-Team Report Error Messaging

We’ve enhanced report error handling to help teams resolve issues quickly:

• Immediate notifications for expired API keys, provider rate limits, or unhealthy connections.
• Clear identification of affected connections and the option to continue, fix, or retry.
• Surfaced unknown errors with raw data for faster troubleshooting.

Red-Team On-Prem (Early Access)

For organizations with strict data residency or air-gapped requirements, Inference Red-Team is now available for early access in on-premise deployments. This option enables private, self-controlled red-teaming with the same capabilities as our SaaS platform.

The Impact of This Release

This release reflects CalypsoAI’s commitment to proactive, enterprise-ready AI security. By unifying detection, prevention, red-teaming, and compliance, our platform secures AI applications and agents at the inference layer. Enterprise-ready integrations like Splunk connectivity and on-prem deployment options ensure that even highly regulated industries can adopt our platform with confidence. Meanwhile, adaptive protections powered by agentic capabilities and scanner versioning, guarantee that our customers’ security posture evolves at the pace of emerging AI threats.