All news in category “AI and Security Pulse”

🤖 AWS now offers Claude Opus 4.8, Anthropic's most capable generally available model, bringing advances in agentic coding, professional knowledge work, and autonomous long-running tasks for developers and enterprises. The model sustains longer sessions, reasons more deeply, and maintains consistency for production workflows. Customers can access Opus 4.8 via Amazon Bedrock or the Claude Platform on AWS, with AWS-managed features and unified billing.

🛡️ CIOs rushing to deploy AI agents without visibility risk major failures; experts warn that observability and governance are required. Many organizations treat agents like RPA and set-and-forget systems, but agents operate in model runtimes and need end-to-end tracing, least-privilege permissions, and human-in-the-loop checks. Vendors and cloud providers offer tools, yet governance can become a bottleneck if it’s not scalable and actionable.

🔍 Cisco researchers found that popular frontier LLMs from OpenAI, Anthropic, Google, xAI, and Amazon exhibit substantially higher risk when subjected to multi-turn adversarial attacks than when assessed with single-prompt safety benchmarks. The team ran tens of thousands of single-turn and multi-turn attacks across 15 models and multiple configurations, revealing wide gaps in attack success rates (ASRs) and configuration-dependent safety behavior. They urge improved benchmarks, transparency on configuration impacts, and publication of paired single- and multi-turn ASRs to better inform procurement and governance decisions.

🧭 This article examines practical strategies to reduce AI cold-start latency on Cloud Run when serving GPU-backed models. It outlines the four-phase cold-start process, highlights storage and model-format choices (Cloud Storage, container images, GGUF, Safetensors, quantization), and explains Cloud Run features like image streaming, temporary CPU boosts, and concurrency tuning. The piece also shares operational tactics—warmup endpoints, startup probe tuning, regional deployment choices—and production patterns used by Elastic to treat GPUs as fungible compute.

🔒 Cisco researchers warn that safety guardrails in several leading large language models (LLMs) can be bypassed through multi-turn conversations. They tested frontier models including ChatGPT, Claude, Gemini, Nova and Grok, finding many were susceptible to manipulation that yields disallowed outputs. Techniques such as roleplay, ambiguity, reframing, and persona adoption were effective, and model configuration affected resilience.

🛡️ Recent developments show AI moving from automation to autonomous cyber operations, shifting how offense and defense interact. The Anthropic Mythos Preview and related incidents illustrate models discovering and chaining vulnerabilities with limited human direction, prompting coordinated defensive responses from major vendors. Policy and procurement are adapting, and security leaders must treat AI agents as principals, invest in adaptive defenses, and reframe risk models for continuous compromise.

🩺 Generative AI chatbots are increasingly used for health questions, but they carry significant risks ranging from incorrect diagnoses to privacy exposures. Users may unknowingly share sensitive medical details that could be used for model training or passed to third parties. Health-focused services vary in their data-handling promises, and most consumer chatbots are not covered by HIPAA. Follow practical precautions and always verify AI advice with qualified medical professionals.

📡 Researchers show WiFi signals can reveal people and environments by analyzing how radio waves reflect, scatter, and absorb compared with expected patterns. WiFi sensing uses these variations to infer spatial structure and presence, effectively creating an image of surroundings and occupants. Thorsten Strufe of KIT explains it functions like a camera, but with radio waves instead of light, enabling recognition through signal propagation analysis.

🚦The author argues that traditional post-hoc compliance reviews fail for AI because AI systems change continuously. Drawing on research into Chinese and EU approaches, the piece recommends embedding governance into CI/CD pipelines so model cards, data lineage and risk evaluations are generated and enforced as deployment gates. It also urges treating agent identity as first-class security control and positioning compliance as operational release infrastructure rather than a review layer.

🔍 Anthropic and over 50 partners report Project Glasswing, using Claude Mythos Preview, has surfaced roughly 10,000 critical or high-severity vulnerabilities across open source projects and partner software. The initiative scanned more than 1,000 open-source projects and validated thousands of findings with independent security firms, but maintainers are overwhelmed by the volume and pace of disclosures. Anthropic is disclosing issues under a coordinated policy and has launched enterprise offerings like Claude Security and a Cyber Verification Program to support legitimate security research.

🛡️ Anthropic appears to be preparing a public rollout of its restricted Mythos model, which the company warned poses major security risks by automating high-quality cyberattacks. Announced in April as an advanced frontier model, Mythos showed dramatic improvements in code reasoning and autonomy compared to Opus 4.7. References briefly appeared in Claude Code and Claude Security, suggesting a controlled preview, while Anthropic builds guardrails and works with partners through its Glasswing initiative.

🛡️ Researchers argue enterprises must stop treating AI agents as trusted components and instead secure them as untrusted systems. The paper, authored by teams from Google, UC San Diego, UW–Madison and others, distills five systems-security principles—least privilege, tamper resistance, complete mediation, secure information flow, and human risk—and maps eleven real-world agent attacks to these violations. They caution that stacking ML guardrails is insufficient and propose research directions for separating instructions from data, verifiable least-privilege policies, and information-flow controls.

🔎 Anthropic disclosed that Project Glasswing and access to Claude Mythos Preview helped partners uncover over 10,000 high- or critical-severity vulnerability candidates across widely used, systemically important software since last month. Analysis verified 1,726 true positives, including 1,094 high- or critical-severity flaws, and resulted in 97 upstream patches and 88 advisories. One notable finding was a critical WolfSSL flaw (CVE-2026-5194).

🦷 Movix built a custom agentic AI platform to address a severe shortage of skilled dental technicians and reduce costly remakes in aligner and appliance manufacturing. Using Google Cloud infrastructure, including Gemini Enterprise Agent Platform, Cloud Run with L4 GPUs, and Compute Engine, Movix developed deep learning, computer vision, and 3D mesh models to automate quality control and data entry. The solution integrates with legacy lab systems, anonymizes PHI for compliance, and targets large-volume labs to improve accuracy, speed, and cost savings.

🔧 Industrial AI initiatives collide with legacy OT realities: an AI-ready control room can still depend on an unpatched Windows 7 maintenance laptop that alone communicates with protection relays. The author reports pervasive visibility gaps across utilities and plants, noting fewer than 10% of OT networks have meaningful monitoring. AI trained on IT telemetry misclassifies normal industrial traffic and automated responses risk shutting down production; passive monitoring of Level 0–2 protocols and a focus on crown-jewel processes are essential before layering AI.

🔒 Microsoft has open-sourced two tools, Rampart and Clarity, intended to embed safety engineering into the AI agent development lifecycle rather than leaving it as a periodic checkpoint. Rampart converts red-team findings into structured, repeatable tests that can be automated in CI/CD pipelines and is built on top of PyRIT for continuous adversarial and benign scenario execution. Clarity targets an earlier phase, guiding engineers through structured conversations to clarify assumptions, expected behaviors, permissions and trust boundaries, storing outcomes as markdown in a .clarity-protocol/ directory for review. Both projects join Microsoft’s broader open-source agent governance stack to address risks such as prompt injection, unsafe tool use, privilege escalation, and unintended autonomous actions.

🛡️ Security professionals at DTX argued that integrating AI into SOCs is now essential to counter autonomous attacker tooling and AI-accelerated threats. Panelists stressed sustaining core cyberdefence fundamentals—system hardening, patching, access control and monitoring—before deploying AI, and preserving human oversight to manage model risk. They noted role shifts toward validation, prompt engineering and GRC, and urged rigorous testing and SDLC-like deployment controls.

🛡️ Microsoft has released two open-source tools, RAMPART and Clarity, to help developers test and clarify AI agent safety early in the development lifecycle. RAMPART is a Pytest-native framework for writing and running adversarial and benign safety tests against agents, building on prior work such as PyRIT. It evaluates test outcomes via simple adapters that connect an agent to the suite, while Clarity acts as a structured thinking partner to surface assumptions, explore failure modes, and guide design decisions before coding begins.

🔒 Microsoft has open-sourced two engineering tools—RAMPART and Clarity—to make agent safety a continuous part of development. RAMPART provides a pytest-style framework that brings red-team and adversarial tests into CI, evaluating tools invoked and side effects. Clarity is a structured design companion that captures problem statements, failure analyses, and decisions in a .clarity-protocol directory. Both aim to create living safety artifacts integrated into normal workflows.

🔒 AI security cannot be reduced to a single benchmark. Over the past 30 years software security evolved from black‑box penetration testing to white‑box analysis and process-driven standards such as BSIMM, and the report argues that AI requires a similar assurance-first approach. Benchmarks fail to capture emergent, systemic properties, so organizations should clean up their WHAT piles, adopt risk-based processes, and accept that there is no simple security meter for AI.