Reports
AI-generated structured vendor updates
Anthropic Claude Opus 5 Goes GA on AWS Bedrock with 0% Prompt Injection
Anthropic launched Claude Opus 5 on AWS Bedrock across 4 regions and on Claude Platform. Auto Mode achieves 0% prompt injection in 129 browser agent tests, refuting OpenAI's claim. Priced at $5/$25 per M tokens, it offers leading performance at half the cost of Fable 5.
Google Cloud Unveils GKE AI Security Blueprint with Three-Layer Defense
Google Cloud launches a security blueprint for AI workloads on GKE, featuring a three-layer defense: Confidential GKE Nodes for hardware memory encryption, open-source k8s-aibom for AI bill of materials, and Model Armor for prompt injection and data leak detection.
MemGhost Attack: Persistent False Memory Injection in AI Agents via Email
Researchers unveil MemGhost, a stealth memory injection attack that plants persistent false memories into AI agents via a single email without user notification. It exploits the persistent memory feature, highlighting critical security gaps and driving demand for memory auditing.
SANS Identifies Distributed Scanning of MCP Servers and AI Assistant Configs
SANS Internet Storm Center reports systematic scanning of MCP servers, AI assistant configs, and local LLM endpoints. 49 IPs targeted MCP handshakes, exploiting CVEs in MCP SDKs, signaling AI infrastructure as a new attack vector.
Cisco AI Defense Adds Agent Harness Red Teaming for Agentic AI Security
Cisco introduces Agent Validation in AI Defense: Explorer Edition, a dedicated red-teaming capability for agentic AI systems. It autonomously probes agent harness attack surfaces, including tool routes, indirect content channels, and persistent state, providing verified findings beyond chat-based security assessments.
Cisco AI Defense + AppOmni Extends Runtime Guardrails to SaaS AI Agents
Cisco integrates AI Defense with AppOmni, using AgentGuard as a real-time intercept layer inside SaaS environments. Custom guardrails now apply to Microsoft 365 Copilot, ServiceNow Now Assist, and other SaaS agents, monitoring MCP, chat, and agent-to-agent channels to block prompt injection, tool exploitation, and data exfiltration with a unified policy engine.
Google AlloyDB Remote MCP Server GA: Standardizing AI Agent Data Access with Open Protocol
Google Cloud announces GA of AlloyDB Remote MCP Server, enabling AI agents to securely access operational data via HTTP endpoints. Built on open MCP protocol, it offers IAM fine-grained authorization, Model Armor protection, and audit logging, integrated with AlloyDB’s ScaNN vector index (10B+ vectors, 6x speed) and AI functions, positioning AlloyDB as the single source of truth for enterprise agentic workloads.
Cisco Provides AI Defense Proactive Testing Platform via DevNet Lab
Cisco released a hands-on DevNet lab for its AI Defense Explorer Edition, enabling developers to conduct agent-driven red teaming on AI models and applications in a self-service manner. The tool uses natural language to define attack objectives and simulates multi-turn adaptive attacks, aiming to shift security testing left in the development lifecycle.
Cisco Extends AI Defense to Google Cloud for Multi-Cloud Runtime Protection
Cisco has extended its AI Defense security platform to Google Cloud, offering runtime protection for AI models, agentic workflows, and RAG pipelines. This move completes its coverage of the three major public clouds (AWS, Azure, Google), aiming to provide a unified multi-cloud AI security framework for enterprises.
Cisco Launches AI Agent Security Scanner, Shifting Security Control Point to IDEs
Cisco has launched an AI Agent Security Scanner IDE extension designed to identify and mitigate new attack surfaces in the AI development toolchain. The tool provides local, multi-layered protection by statically scanning MCP server configurations and agent skill definitions, embedding secure coding rules during code generation, and continuously monitoring file integrity at runtime.
Google Cloud Next '26: Agent Gateway Seizes Control Plane, TPU 8i Locks Inference
Google Cloud Next '26 announces 8th-gen TPUs (8t for training, 8i for inference), Agent Platform with Agent Gateway, Agent Identity, Agent-to-Agent Orchestration, Agentic Data Cloud, and Agentic Defense integrating Wiz. The move shifts control from infrastructure to agent orchestration, locking enterprises into a vertically integrated stack.
Anthropic Launches Claude Opus 4.7 with Cyber Safeguards
Anthropic has launched Claude Opus 4.7, showing notable gains in advanced software engineering, multimodal understanding, and long-horizon reasoning. This release introduces automated safeguards to detect and block prohibited high-risk cybersecurity uses, alongside a Cyber Verification Program for legitimate research, aiming to inform the safe future release of more powerful models like Mythos.
Cisco Research Uncovers New Multimodal Prompt Injection Risks and Defense Signals
Cisco's AI security research team published a report systematically assessing typographic prompt injection attacks against Vision-Language Models. The study found that visual transformations like font size, blur, and rotation significantly impact attack success rates. It also proposes text-image embedding distance as a lightweight, model-agnostic signal for flagging risky inputs, offering a new approach for building multimodal AI security defenses.
Cisco RSA 2026: Three-Pillar Security Framework for Agentic Workforce
At RSA Conference 2026, Cisco unveiled a three-pillar security framework for agentic workforce: 1) Zero Trust for AI Agents - Duo IAM integration with MCP policies for verified agent identities; 2) AI Defense Explorer Edition - Dynamic red teaming tool supporting prompt injection and jailbreak simulations; 3) Splunk SOC enhancements - Exposure Analytics, Detection Studio, Agentic SOC Expansion with specialized agents like Detection Builder and Triage Agent. Also launched DefenseClaw security framework and LLM Security Leaderboard.
Cisco Launches Open-Source AI Agent Security Solution DefenseClaw
Cisco released open-source security solution DefenseClaw with four protection engines for OpenClaw AI Agent, covering prompt inspection, tool detection, installation scanning and code review. The solution demonstrates defense against 11.9% identified threats including malicious skills and unsafe MCP servers through hands-on labs.
Cisco Open Sources DefenseClaw for AI Agent Security Governance
Cisco launched open-source DefenseClaw, providing three-layer security architecture for AI agents like OpenClaw: supply chain scanning, runtime inspection, and system boundary control. The solution integrates NVIDIA's OpenShell sandbox for end-to-end automated governance.
Cisco Extends Zero Trust Security to AI Agent Ecosystem
At RSA 2026, Cisco introduced security innovations for AI agents, extending Zero Trust Access with agent discovery in Identity Intelligence, agentic IAM in Duo, and MCP enforcement in Secure Access SSE. It launched AI Defense: Explorer Edition for self-serve testing and DefenseClaw open source framework to automate security deployment.
Check Point AI Factory Blueprint: Security Control Shifts to NVIDIA DPU and LLM Layer
Check Point unveils AI Factory Security Blueprint, tightly integrating its firewall with NVIDIA BlueField DPU via DOCA. The architecture enforces security at four layers: LLM, AI infrastructure, perimeter, and workload. The new AI Factory Firewall delivers hardware-accelerated threat prevention without consuming CPU/GPU cycles, aiming to embed security into the AI fabric.
OpenAI Proposes Architectural Defense for AI Agents Against Prompt Injection
OpenAI released a technical approach implementing action constraints and data protection in AI agent workflows to defend against prompt injection attacks. The solution restricts API access and isolates sensitive data, shifting from passive filtering to proactive security architecture.
Cisco Elevates Prompt Injection Defense to Infrastructure Layer
Cisco compares prompt injection to SQL injection, advocating layered defense including network micro-segmentation and EDR-based endpoint protection to mitigate LLM security risks.