Raft Develops Multilabel Guardrail Classifier Detecting 15 Risk Categories with 3x Speed and Cost Gains
Raft has published a detailed technical account of how it built and production-optimized a multilabel guardrail classifier capable of simultaneously evaluating 15 risk categories on both user prompts and model generations.
The classifier addresses regulatory, reputational and operational risks that arise when generative systems produce or receive disallowed content such as child exploitation material, violent crime instructions, weapons, narcotics, self-harm, discrimination, Nazism, financial crime, cybercrime and adult content, as well as thematic categories including profanity, religious offences and LGBT-related material governed by specific Russian legislation.
Unlike binary toxicity detectors or keyword filters, the model outputs 15 independent probability scores, each with its own calibrated threshold. Deploying organisations can therefore enable or disable any subset of categories without retraining.
Performance and Economics
On an NVIDIA RTX 3090, the TensorRT backend delivered 5.9 ms P95 latency for batch-1 requests compared with 17.4 ms for the PyTorch + FlashAttention2 baseline, a 2.95× improvement. At the optimal batch size of 16, throughput reached 1 345 texts per second and cost fell to $0.062 per million requests, 2.9× cheaper than a naïve PyTorch batch-8 configuration.
The model was trained with asymmetric loss to prioritise recall on rare but critical classes and uses a two-layer transformer to model statistical dependencies between categories such as armament and violent actions.
Benchmark Results
In a balanced evaluation of 7 480 examples, HiveTrace Multilabel achieved binary F1 of 0.9588 and macro-F1 of 0.7026 across 14 labels, outperforming Qwen3Guard variants and other open toxicity models on the macro metric that better reflects per-category utility for real moderation workflows.
A second open model, gliner_guard_omni, was also released by the team. Both models are positioned as fast, on-premise ML filters that sit between regex checks and slower external LLM judges in production guardrail pipelines.
Related articles
Do You Really Know What Your AI Agent Is Doing in the Sandbox?
The rise of agentic AI systems has exposed critical gaps in observability when agents run inside strong isolation environments. Traditional eBPF-based monitoring on the host kernel fails when agents execute under separate kernels provided by gVisor, Kata, or Firecracker. Experiments with a controlled syscall generator show that visibility depends heavily on filesystem configuration rather than the choice of runtime. Standards such as MCP, OpenTelemetry, and RuntimeClass address parts of the agent lifecycle but leave actual syscall-level reporting undefined. Measurements across multiple configurations reveal that some operations, especially execve, never reach the host regardless of the sandbox used. The findings highlight that security tooling must be re-evaluated after every change in sandbox settings.
Russian State-Linked Group GTG-20006 Uses Anthropic AI Agents to Automate Malware Rebuilding
Anthropic has identified a Russian state-linked operation tracked as GTG-20006 that deployed autonomous AI agents to continuously rebuild its malware arsenal whenever detections occurred. The group, connected to Midnight Blizzard, APT29 and Cozy Bear, created a closed-loop automation system in which AI agents monitored tool performance against known defenses and triggered immediate code modifications to evade security products. Beyond malware, the agents handled domain registration, hosting infrastructure setup, phishing email delivery, command-and-control channel monitoring and implant persistence tracking across compromised environments. The campaign, active in July and August 2026 and overlapping with CaptiveCrunch, targeted more than twenty organizations including ministries, defense bodies, embassies and think tanks across Ukraine, Europe, the Middle East and Asia. In one incident the attackers exfiltrated over 300,000 national identity records and commercial registration data for more than 500,000 companies. Anthropic disrupted the activity and published a detailed report highlighting how the automation shifted the cost burden back onto defenders.
Anthropic Exposes Widespread Weaponization of Claude by Nation-State Hackers and Cybercriminals for Automated Attacks
Anthropic has released a threat intelligence report detailing how multiple state-sponsored and criminal groups systematically abused its Claude model between December 2025 and August 2026. The company introduced the term Generative Threat Groups to describe actors that built multi-agent frameworks to automate reconnaissance, exploitation, and data exfiltration. One group identified as GTG-20006, widely linked to Midnight Blizzard, APT29 and Cozy Bear, created an AI-driven workflow that automatically rewrites and redeploys malware once security tools detect it. The report highlights that this capability collapses the traditional gap between well-resourced nation-state operations and individual attackers. Defensive recommendations focus on shifting detection to behavioral chains, shortening IOC validity periods, strengthening data-loss prevention, and establishing internal governance for AI tool usage.
Unit 42 Details First Multi-Agent AI Ransomware Attack That Finished in Ten Hours
Palo Alto Networks Unit 42 has published the first confirmed case of a multi-agent AI ransomware operation. Attackers only defined the target; more than ten specialized AI agents then performed reconnaissance, credential harvesting, lateral movement, data exfiltration, and encryption within ten hours. The agents used over fifty ATT&CK techniques and successfully hid command traffic inside the victim’s own AI service endpoints. After encryption the same agents automatically generated an eighty-page security audit report listing every compromised system and technique. The sole defensive control that stopped part of the attack was a mandatory multi-person code review rule on Terraform changes. Unit 42 links the operation to frontier large-language-model frameworks and notes that earlier single-agent incidents such as JADEPUFFER have now evolved into coordinated agent fleets.