Cyber Brief: Patch pressure, Fortinet exposure and DDoS disruption

Today’s cybersecurity picture is a reminder that operational risk builds from several directions at once. This week brings a large Microsoft patch cycle, newly published critical Fortinet flaws, fresh evidence that trust assumptions in AI assisted workflows are still too weak, and another major law enforcement move against DDoS for hire services. For UK businesses, the common thread is straightforward, security works best when ownership is clear and controls are practical enough to hold up under pressure.

Microsoft’s April updates put patch governance and user safeguards back in focus

Microsoft’s April Patch Tuesday fixed 165 new CVEs, including CVE 2026 32201, a SharePoint Server spoofing flaw that was already being exploited before a fix was issued. The Register reports that the vulnerability could let an attacker manipulate how information is presented to users, creating a route for phishing, social engineering, or unauthorised changes to disclosed information. At the same time, Microsoft’s April Windows updates added stronger protections for Remote Desktop files, so users now see requested connection settings before connecting and those settings are turned off by default.

For most organisations, this is about more than monthly patching. It is about whether internet facing systems, collaboration platforms, and routine user workflows are covered by a response process that can move quickly when exploitation or abuse is already in play. The added RDP protections are a useful example of why user level safeguards still matter alongside vulnerability remediation.

Why it matters

For UK businesses running Microsoft estates, this is a good point to confirm both urgent patching and the user experience around remote access. Good security posture is not only about applying fixes, it is also about making risky actions easier to spot before users click through them.

Fortinet’s latest FortiSandbox advisories deserve prompt review

Fortinet published two critical FortiSandbox advisories on April 14. One, CVE 2026 39808, is an OS command injection issue that may allow an unauthenticated attacker to execute unauthorised code or commands via crafted HTTP requests. The other, CVE 2026 39813, is a path traversal issue in the JRPC API that may allow an unauthenticated attacker to bypass authentication and escalate privilege. Fortinet says affected versions include FortiSandbox 4.4.0 through 4.4.8, with fixes available in 4.4.9 and later, while FortiSandbox 5.0.0 through 5.0.5 is affected by the authentication bypass issue and should be upgraded to 5.0.6 or above.

This is the sort of vendor advisory that can easily get lost in a busy week, yet sandboxing products often sit close to malware analysis, email security, and wider detection workflows. That makes them important infrastructure, not side systems. Where security tooling itself is exposed or behind on updates, it can create a gap in exactly the part of the environment meant to improve resilience.

Why it matters

If your organisation uses FortiSandbox, this should move quickly into validation and change planning. Reviewing exposure, upgrade paths, and who owns security appliance maintenance is often more useful than treating these advisories as purely technical notices.

New AI agent findings show that trust models are still too weak

Two fresh reports this week underline a wider issue in AI assisted engineering. The Register reported on Ox Security research describing several classes of weakness tied to the Model Context Protocol, including zero click prompt injection across AI coding tools such as Windsurf, Claude Code, Cursor, Gemini CLI, and GitHub Copilot. In a separate report, The Register covered Manifold Security research showing that an AI powered code reviewer built on Claude could be tricked into approving malicious code by spoofing trusted Git author metadata with only two Git commands.

The operational lesson is not that teams should stop using AI in development. It is that these systems should not inherit trust from surrounding context without stronger verification. Tool permissions, commit signing, reviewer design, prompt boundaries, and the handling of untrusted external content all need more attention before AI review and automation are treated as routine controls.

Why it matters

For businesses adopting AI in software delivery, governance now needs to catch up with usage. It makes sense to review where AI tools are allowed to act, what trust signals they rely on, and whether those signals can be forged or influenced too easily.

Operation PowerOFF is another reminder that DDoS remains accessible and disruptive

CyberScoop reported on the latest phase of Operation PowerOFF, in which authorities from 21 countries took down 53 domains, arrested four people, and obtained data on more than 3 million alleged criminal user accounts linked to DDoS for hire services. Europol said more than 75,000 users engaged in DDoS attacks were targeted by the operation. The agencies also issued warnings to participants, removed more than 100 URLs from search results, and used search ads to reach people looking for these services.

For defenders, the point is not only that law enforcement action is continuing. It is that DDoS capability remains easy to access and is often used against web services, marketplaces, telecommunications providers, and other internet facing businesses. Resilience planning still needs to cover availability, not just compromise and data loss.

Why it matters

For many organisations, DDoS preparation is best treated as part of normal service resilience. Reviewing mitigation arrangements, supplier dependencies, escalation paths, and communications plans can make a meaningful difference when disruption hits customer facing services.

Today’s Key Actions

• Confirm whether Microsoft’s April updates have been applied across SharePoint, Windows endpoints, and other priority systems, then check whether users who open RDP files will now see the stronger warnings and settings review.
• Review any FortiSandbox deployments, validate affected versions, and move upgrade actions into the current change queue where needed.
• Audit AI enabled development and review tooling for trust assumptions around commit metadata, external inputs, and connected tools or agents.
• Check whether DDoS response planning is current for customer facing services, including supplier contacts, mitigation arrangements, and internal ownership.
• Make sure patching, appliance maintenance, AI workflow governance, and service resilience each have clear accountability so urgent decisions do not stall.

Secarma Insight

This week’s themes show why mature security is built on dependable foundations. Fast patching, sound configuration control, sensible trust models, and tested resilience planning all depend on the same basics, visibility, ownership, and the ability to act without confusion when priorities shift. Organisations that handle weeks like this well are usually the ones that have already turned security into a working discipline across technology and operations.

That is where Secarma helps, supporting businesses with clear advice, practical testing, and assurance that strengthens security posture while supporting growth.