🤖 Managing Agentic Security at Scale

Salesforce's two-track remediation model for AI-discovered vulnerabilities: a strategic track fixes shared architectural choke points, retiring findings in bulk; a tactical track closes validated multi-step exploit chains on adversary timelines.

https://www.salesforce.com/blog/managing-agentic-security-at-scale

#AI

🤖 Getting Agents to tell on themselves

Thinkst's Agent Provocateur is a Canary personality that exploits agent suggestibility. It greets scanning agents, engages them helpfully, and probes for username, MAC address, and SSH host-key fingerprint, raising alerts. Internal tests reported agents tripping it 100% of the time.

https://blog.thinkst.com/2026/09/getting-agents-to-tell-on-themselves.html

#AI

🤖 How the Claude Code team does code review

Claude Code's team automates most code review via pre-PR verification, AI and security reviewers that block merges, risk-tiered human approval, merge queues, feature flags, and fast rollback. Humans review risky code and architecture, minimizing time without raising reverts or incidents.

https://claude.dev/blog/solving-code-review-at-anthropic

#AI

🤖 Your AI agent fixed the bug, but what else did it change?

Trail of Bits released patching-bugs and regression-hunt, agent plugins for reproducing bugs, developing fixes, and detecting regressions via live PoCs. regression-hunt yields 25% fewer false positives than a baseline Codex prompt, but confidently reports fewer known regressions.

https://blog.trailofbits.com/2026/10/21/your-ai-agent-fixed-the-bug-but-what-else-did-it-change

#AI

Kube API не отвечает, LoadBalancer не пропускает трафик, DNS не резолвит, исходящие запросы не уходят. Причина обычно кроется на стыке облачной сети и кластера, и найти её без понимания всей цепочки сложно.

3 ноября в 12:00 Роман Ерема, разработчик в команде Managed Kubernetes MWS Cloud Platform, на практике покажет:
⏺️как взаимодействуют сеть облака, CNI и сеть Kubernetes
⏺️как работают SNAT и Source IP
⏺️когда нужен externalTrafficPolicy: Local и почему при нём может не проходить health check
⏺️как диагностировать проблемы на реальных кейсах (live-demo)

Все участники получат чек-лист для проверки доступности.

Полезно инженерам, разработчикам и специалистам по сетям и облачной инфраструктуре.

🗓 3 ноября в 12:00

Регистрируйтесь по ссылке

🤖 I Gave 14 LLMs AWS Keys. Here’s What They Wanted to Do Next

What models propose after a cloud compromise, and the detections we can build from it.

https://medium.com/@adan.alvarez/i-gave-14-llms-aws-keys-heres-what-they-wanted-to-do-next-48585761a11b

#AI

🔴 Introducing new scalable perimeter options with VPC Service Controls

Google Cloud added folder membership to VPC Service Controls perimeters, so contained projects inherit policy. Ingress/egress rules now accept folders, organizations, and Private Service Connect endpoints.

https://cloud.google.com/blog/products/identity-security/introducing-new-scalable-perimeter-options-with-vpc-service-controls

#gcp

🤖 AI in cyber security problems

NCSC chapter on AI in cyber security covers autonomous cyber defence, AI system security, and national-scale resilience. Agentic AI offers defenders scale but adds risks, as threat actors already use AI.

https://www.ncsc.gov.uk/collection/problem-book/ai-cyber

#AI

🔴 Inside Agent Compute Engine: Building the compute foundation for the agentic era

Agent Compute Engine gives each AI agent a hardware-isolated microVM with container-like speed and no in-sandbox credentials. An external Agent Gateway swaps in credentials, idle agents are checkpointed, and memory is deduplicated via DAX and copy-on-write.

https://cloud.google.com/blog/topics/ai-infrastructure/agent-compute-engine-deep-dive

#gcp

🤖 Cross-request KV cache exposure in the GLM-5.3 Workers AI model, and our remediation with SGLang

A configuration mismatch in GLM 5.3 on Workers AI triggered an SGLang bug that exposed KV cache pages across requests. Here is how Cloudflare addressed the issue.

https://blog.cloudflare.com/kv-cache-exposure-remediation

#AI

🚨 Broccoli2shell: Everything You Need To Know About The Critical Brotli Decoder Vulnerability

Detect and mitigate CVE-2026 111497, a critical Brotli decoder vulnerability enables remote code execution.

https://www.wiz.io/blog/cve-2026-111497-brotli-rce-node-js

#CVE

🤖 Securing our frontier model training environment

OpenAI's post which describes how they secure the environments used to train and evaluate frontier models.

https://openai.com/ru-RU/index/securing-our-frontier-model-training-environment

#AI