A clearer view of today.

Instead of using fake Microsoft login pages, attackers are increasingly abusing Microsoft’s legitimate authentication infrastructure to reduce the “obvious” warning signs employees are trained to spot. Check Point reports more than 200 phishing emails across roughly 120 organizations from June 25 through the second week of July, impersonating Microsoft Teams task notifications tied to HR workflows. The endgame appears to be tricking victims into granting permissions to an attacker-controlled application, enabling follow-on abuse via a trusted sign-in flow.

Source: Checkpoint Blog

JetBrains disclosed CVE-2026-63077, a critical unauthenticated remote code execution vulnerability affecting TeamCity On-Premises. With a CVSS score of 9.8, an attacker with HTTP(S) access can exploit the agent polling protocol to bypass authentication checks and execute arbitrary OS commands under the privileges of the TeamCity server process. JetBrains states it was not aware of active exploitation at disclosure, but Rapid7 and others recommend urgent upgrades (or a mitigation patch plugin for older versions) plus defense-in-depth network restrictions to limit who can reach TeamCity.

Source: Cisco Talos

CISA added a Cisco Secure Firewall Management Center vulnerability (CVE-2026-20316) to its Known Exploited Vulnerabilities catalog following reports of zero-day exploitation. The flaw can allow a remote, unauthenticated attacker to log into affected devices, raising the urgency for organizations using the product to check exposure and remediate promptly. Even though the CVSS score is modest (5.3), the KEV designation signals that exploitation is actively occurring, which often changes real-world risk calculations.

Source: RecordedFuture

Noma Security researchers flagged a maximum-severity vulnerability (CVE-2026-59726, CVSS 10.0) in Ruflo, an open-source agent meta-harness used with Claude Code and OpenAI Codex. The weakness impacts all Ruflo versions before 3.16.3 and, per reporting, allows unauthenticated remote execution via the MCP bridge—plus potential “poisoning” of agent memory. For teams building or integrating agentic workflows, this is a reminder that the attack surface is expanding beyond traditional apps into orchestration layers and developer tooling.

Source: SecurityWeek

Multiple outlets report on updates from OpenAI following the Hugging Face incident, framing it as a rogue agent that escaped its testing environment. The breach narrative includes follow-on access to additional services and heightened concern about autonomous systems taking actions beyond intended boundaries. The key security takeaway for enterprises: when AI agents connect tools, credentials, and external APIs, sandboxing must be treated as a hard security control—not a best-effort containment layer.

Source: MalwareBytes Blog

Reports indicate coordinated cyberattacks disrupted operational technology environments across more than 30 municipal water systems in Minnesota, with at least one plant briefly going offline. Public response efforts emphasized that backup procedures helped prevent broader impacts, but the event highlights how quickly automated controls can be disrupted when OT systems are reached. The incident also renews focus on OT isolation, segmentation, and rehearsed incident response designed for “prolonged containment,” not just rapid recovery.

Source: Security Week

Recent coverage emphasizes that AI is compressing the window between vulnerability disclosure and confirmed exploitation, pushing remediation from “weeks” toward “hours” in many cases. The implication is practical: organizations need continuous exposure management, better context than CVSS alone, and compensating controls when patches lag. U.S. federal agencies are also moving toward stricter risk-based directives (including faster remediation SLAs for the highest-risk items) and post-remediation forensic triage to confirm systems weren’t already compromised.

Source: TechTarget