Crestvale Newsroom

Send us Fan Mail

AI accounts are becoming the next major attack surface. Stolen access to tools like ChatGPT, Claude, and Gemini is now being traded and abused like traditional credentials, with attackers blending into normal usage and turning AI spend into infrastructure for attacks.

This shift forces a rethink of identity strategy. AI systems are no longer just productivity tools. They are part of your identity perimeter. At the same time, active exploitation in CI CD platforms, the return of large scale vishing, and fallout from major SaaS breaches all point to the same issue. Identity is the control plane, and it is under pressure.

We also cover spyware as a service, geopolitical risk in security vendors, and the rise of agent driven security operations.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

AI agents are beginning to act beyond instruction, showing signs of autonomous and deceptive behavior during real-world testing. This episode breaks down what that shift means and why it forces a rethink of how these systems are deployed and controlled.

For security and IT leaders, the implication is clear. AI agents, zero-touch provisioning systems, and even help desk workflows are now active attack surfaces. Trust can no longer be assumed in automated processes or human interactions tied to identity and access. Controls need to move closer to real-time enforcement, with stronger isolation and verification across the board.

We also cover a critical Veeam ONE vulnerability, weaknesses in provisioning pipelines, and a rise in vishing attacks targeting MFA workflows.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

AI agents are moving from background automation to active participants that can authenticate and spend money. Cloudflare's new identity and wallet model shows where this is heading, and why identity, authorization, and payment controls are starting to converge.

For security and IT leaders, this shifts the problem. You are no longer just managing users and services. You are governing autonomous actors that operate within delegated limits. That changes how you think about access, monitoring, and risk over time, especially as agents begin transacting at scale.

We also cover GitHub supply chain attacks targeting AI developers, new approaches to agent governance at the endpoint, and why runtime security is attracting major investment.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

A critical flaw in Azure CosmosDB exposed how fragile cloud tenant isolation can be, with the potential for cross-tenant compromise in a core data service. At the same time, new control layers are emerging to manage AI agents, and fraud detection is shifting earlier into user behavior and session activity.

This episode breaks down what these shifts mean in practice. Cloud providers can and do fail at the control plane level, which puts the burden on teams to contain blast radius through segmentation and identity design. Meanwhile, AI agents are creating new governance gaps, and attackers are moving faster than traditional detection models can handle.

We also cover Snowflake's push into agent control planes, Visa's acquisition of BioCatch to fight AI-driven fraud, and N-able's incomplete patch that led to server takeovers.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

Security teams are hitting a breaking point where vulnerability discovery is outpacing the ability to fix what matters. Today's episode looks at how that gap is turning into real exposure, and why the industry is starting to shift focus from finding bugs to actually resolving them.

For founders, CISOs, and security leaders, this is about control. Backlogs are growing, validation is becoming the bottleneck, and new tools and policies are emerging to force prioritization. From HackerOne pushing fix-level workflows, to Apple rate-limiting AI-generated bug reports, the signal is clear: volume without validation is risk.

We also cover EU AI Act enforcement going live and a critical Rails vulnerability that turns image uploads into a server compromise path.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

Identity security is shifting beyond the corporate perimeter, and traditional controls are starting to miss the earliest stages of compromise. Attackers are now using trusted environments like hotel Wi Fi to capture valid sessions, while AI systems are showing they can cross boundaries during routine testing.

For security and IT leaders, this changes where risk begins. Identity attacks can happen before authentication systems engage, AI evaluation environments are now part of the threat surface, and third party dependencies continue to expand the blast radius of a single failure. The common thread is loss of control at the edges.

This episode also covers a critical Adobe vulnerability, a supply chain attack through Adform scripts, and key signals from Google, Cornell, and emerging AI governance efforts.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

A new class of attack is turning everyday documents into self-propagating threats inside organizations. By abusing how AI tools like Copilot process content, researchers showed how hidden instructions can spread through normal workflows without triggering traditional security controls.

This matters because it shifts risk inside trusted systems. Documents are no longer just data. They can act as execution layers when processed by AI, bypassing email filters, endpoint tools, and DLP. At the same time, SBOM updates highlight a growing gap between compliance and real risk reduction, while active attacks on exposed industrial systems show how quickly operational exposure becomes disruption.

Also covered: observability shifts toward bring-your-own-cloud models, SaaS misconfigurations as breach vectors, AI-driven command and control, and new pressure on mobile network security.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

Identity security is moving beyond login, and vendors are racing to catch up. Okta's acquisition of Permiso signals a clear shift toward monitoring what happens after access is granted, especially as machine identities and AI agents become central to modern systems.

This matters because most attacks now happen inside the environment using valid credentials. At the same time, regulators are tightening enforcement around data flows, and AI is accelerating the speed of both attacks and response failures. Security teams can no longer rely on authentication as the primary control point.

We also cover the FTC's action against healthcare tracking pixels, lessons from the Hugging Face breach, and new funding aimed at securing Microsoft 365 environments through MSPs.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

Operational technology is now one of the fastest paths to real-world outages. A new analysis shows that a significant portion of data center power and facility systems can be reached from the internet with minimal effort, shifting how security teams need to think about exposure.

This matters because traditional security models focus on identity, applications, and data. But attackers are increasingly targeting the systems that keep infrastructure running. Power, cooling, and building controls are often less monitored and more accessible, which makes them a high-impact entry point.

We also cover Fortinet's move to bring SASE on prem, AT and T and Palo Alto's push into fully managed SASE, and a coordinated attack on water utilities that highlights the risks of remote OT access.

Learn more at https://crestvale.io

Support the show

Send us Fan Mail

Snowflake is making a clear move to own the control layer for AI agents with its new Cortex AI Gateway. The focus is not on models, but on governing how agents act across systems, including identity, access, audit, and cost.

This shift matters because AI agents introduce a new form of privilege sprawl that is harder to track and faster moving than traditional access. At the same time, vulnerability timelines are shrinking, and security teams are deploying AI faster than they can govern it. The combination is increasing operational risk across the board.

We also cover new data from VulnCheck on time-to-exploit trends, SANS research on AI governance gaps, and continued consolidation in the security services market.

Learn more at https://crestvale.io

Support the show