TLDR IT 2026-10-28

IT leaders are under pressure. They need to scale AI faster than the competition or risk losing out – but they must also protect the security, data, and governance boundaries that AI exposes to attacks, leaks, and other breaches.

Atlassian's upcoming demo (Oct 29th) will reveal how to give teams AI that understands their organization's context and respects who can see what. You'll learn from end-to-end workflows for different teams:

• Accelerate incident response timelines in Engineering and IT.
• Reduce repeat questions + speed up onboarding in HR.
• Build accurate briefs, perform effective AI drafting & async approvals in Marketing.

Save your virtual seat + get the recording.

SAP has unveiled its physical AI capabilities designed to connect enterprise business systems with physical operations such as robots and automated systems. At SAP TechEd, the company showcased implementations with partners like Foxconn, Arçelik-LG, Intel, and Holcim to demonstrate how business context can govern physical tasks while automatically feeding structured execution data back into enterprise workflows.

A patched flaw in Microsoft Authenticator for Android let malicious apps impersonate trusted Microsoft apps and obtain work-account tokens, exposing email, files, and calendars in researchers' tests. Microsoft says exploitation requires a user to complete a sign-in flow and reports no known attacks. The fix is available in version 6.2606.4246 and later.

As recurring vulnerabilities expose legacy VPN gateways to rapid exploitation, organizations will need to transition to zero-trust architecture. Traditional VPNs grant broad internal network access once breached, a risk amplified by AI-driven vulnerability discovery. Zero-trust models replace this perimeter approach by making internal systems invisible to the public internet and restricting users and agents to specific applications.

Linux servers can show green in an EDR dashboard while collecting little or no security telemetry because of kernel incompatibilities or degraded sensor modes. Verify sensor health on each host before expanding coverage or switching vendors, then pilot changes through kernel updates and roll them out in stages. Container visibility and after-hours alert response also need explicit coverage.

Code assistants like Cursor and Copilot are now in ~1/3 environments Field Effect protects. Unlike chat tools, they execute: downloading files and running scripts on developer endpoints with elevated permissions. Field Effect's State of AI Adoption eBook covers the implications. Read the findings, plus more insights into agents and on-device models: download the eBook.

IBM announced new capabilities at TechXchange 2026 to provide enterprise security and governance for AI agents, combining verified identities, access controls, and performance evaluations. The updates extend HashiCorp Vault to issue short-lived credentials for non-human identities, integrate with NVIDIA OpenShell, and use watsonx Orchestrate to discover, contain, and evaluate agents across third-party platforms.

Oracle is expanding its Autonomous AI Lakehouse with a Business Semantic Layer and a Business Simulation Engine designed to ground AI agents in enterprise data and improve analytical decision-making. The company also announced expanded Apache Iceberg support to simplify data management using SQL. The new capabilities are intended to help organizations scale agentic applications and improve data consistency.

SAP announced plans to deploy its tabular AI foundation model, SAP-RPT, on the HPE Spaceborne Computer-2 aboard the International Space Station. The deployment aims to test real-time edge telemetry analysis, anomaly triage, and predictive maintenance in space before applying the same structured-data models to terrestrial supply chains and enterprise operations by early 2027.

Cisco has introduced Privileged Infrastructure Access, a new offering designed to provide network administrators and automated agents with secure, identity-bound connections to critical network devices. The system replaces long-lived credentials and static access paths with short-lived, task-specific privileges that expire once work is completed, reducing vulnerabilities that can be exploited by AI-powered adversaries.

Tailscale is preparing patches for SSH flaws that could let authorized users exceed their assigned privileges, advising customers to tighten access policies and disable unnecessary SSH forwarding in the meantime.

Researchers recently discovered critical vulnerabilities in the PHP programming language that exposed a vast range of enterprise products and websites to potential attacks.