Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Today, we are extending Cloudflare’s cloud access security broker (CASB) to support the Claude Compliance API. Security and compliance teams can now monitor Claude usage directly in the Cloudflare dashboard. No endpoint agents required. Enterprise security teams have long struggled to see how users interact with sanctioned and unsanctioned applications. The rapid adoption of AI applications has made this harder.

In July 2025, Replit's autonomous AI coding agent deleted a live production database despite being explicitly instructed to freeze all changes. The agent then attempted to reassure the user with incorrect information after the fact. The team had safeguards in place. The instructions were explicit. Neither stopped it. The conclusion that follows is one the security community should take seriously: you cannot enforce AI agent behavior through the agent itself.

See how Torq harnesses AI in your SOC to detect, prioritize, and respond to threats faster. Request a Demo Noam Cohen is a serial entrepreneur building seriously cool data and AI companies since 2018. Noam’s insights are informed by a unique combination of data, product, and AI expertise — with a background that includes winning the Israel Defense Prize for his work in leveraging data to predict terror attacks.

Artificial intelligence is rapidly reshaping how security operations centers (SOCs) function. Many organizations are now evaluating AI-native architectures to reduce workload and accelerate investigations. A new architectural narrative is emerging. A growing set of AI-native security vendors are proposing centralizing telemetry in a warehouse and deploying AI agents to replace the operational role of the SIEM. They want to centralize telemetry, apply AI, and automate the SOC.

Our previous post, How to Secure Microservices with SPIFFE and Istio, showed how to secure Kubernetes microservices using Istio policy and SPIFFE identities, with Teleport issuing the identities that the mesh trusts. The question teams face next is: How do you extend that identity-driven security model to workloads outside Kubernetes — such as VMs, edge gateways, and legacy services — without creating a massive certificate-management project?

Snyk started as a classic product-led growth company. For our first two years, we didn't need a sales team — the product sold itself to developers. That's a rare thing, and we're proud of it. It meant we had genuine product-market fit before we had a go-to-market motion. But markets evolve, and so did we. Today, AI coding agents are generating code at a velocity that significantly outpaces the ability of security teams to review it.

Today, we're announcing two new integrations with Anthropic that cover both sides of AI-assisted development. Evo by Snyk now integrates with Anthropic's Claude Enterprise, giving security and compliance teams a complete inventory of their Claude environment models, approved MCP servers, per model risk signals, and tool-level permissions in the platform they already use to govern the rest of the stack.

A phishing campaign exploited a glitch in Robinhood’s account creation process to send phishing emails from the investment platform’s own systems, SecurityWeek reports.

GitHub disclosed that attackers accessed its internal repositories after compromising an employee device through a poisoned Visual Studio Code extension. The company said the activity appears limited to GitHub-owned internal repositories, with the attacker’s claim of roughly 3,800 repositories being “directionally consistent” with its investigation. GitHub also said it found no evidence that customers’ own enterprises, organizations or repositories were impacted.