Every penetration testing firm is facing the same pressure right now. AI tools are faster, cheaper, andincreasingly capable, and testers are using them whether the company has a policy on it or not.

There’s ahigh change AI has already entered your workflow the question is whether it has entered on your terms oryour employee’s personal Claude account.And in this industry, that distinction really matters.

The Cloud AI Problem No One Wants to Talk About

When a developer pastes code into ChatGPT, the worst case is a licensing headache. When a penetration tester does the same thing with a client’s findings, you’ve just uploaded a detailed map of a company’s vulnerabilities to a third-party cloud infrastructure you have no control over.

Besides a policy violation, that’s a potential breach of client confidentiality, NDA, and in regulated industries, compliance frameworks like GDPR or HIPAA.

The major AI providers (OpenAI, Anthropic, Google) are excellent products. But their cloud-hosted models were not designed with the data sensitivity of offensive security work in mind.

Client scope documents, discovered vulnerabilities, network diagrams, hardcoded credentials found during testing. None of that should be leaving your infrastructure, and none of it should be in a training pipeline you didn’t consent to.

The truth is that if your firm doesn’t give testers a sanctioned AI workflow, they will build their own. And it’s not likely they’ll be thinking about your client contracts when they do.

The Right Architecture Exists

The answer isn’t to ban AI. That ship has sailed, and the firms that try will just end up with testers using it covertly while competitors who embrace it properly pull ahead on speed and output quality.

The answer is to bring AI in-house, on infrastructure you control, with client data that never leaves your network.

Self-hosted AI models have matured significantly. Running a capable model locally is no longer just a research project, but an operational choice.

And when you combine a self-hosted model with an orchestration layer built specifically for security workflows, you get the productivity benefits of AI without handing your clients’ attack surface to a third party.

AI specifically designed for cybersecurity companies

PentestPad is a pentest reporting tool that reduces the time security teams spend turning raw notes into a client-ready report, while keeping the project organised around it.

The teams using it are not all in the same business. Some deliver security testing to clients. Some test their own systems in-house.

Others coordinate several external providers at once, and use the platform to keep the findings in one place.

Day to day that means project management, collaboration, retesting and report generation. Much of the tooling that usually surrounds a pentest sits inside it as well: notes, evidence and attachments, comments on findings, proposals and scope import.

Less of the work needs a scratch file, a shared drive or a spreadsheet.

Almost none of it is fixed. Reports generate from your own template rather than a generic one, the client portal carries your branding on your own domain, and the fields, layouts and statuses behind a finding are defined by your team rather than inherited.

The instructions the AI assistant writes to are yours as well, so a draft comes back in your methodology and your wording rather than a generic house style.

The same holds for where it all runs. PentestPad is available as a managed cloud service or fully self-hosted on your own infrastructure, including air-gapped, and the writing assistant can be pointed at a self-hosted model inside your own network rather than a vendor’s.

That last choice is the one pentest teams get stuck on. A finding is a client’s unremediated vulnerability, and handing it to OpenAI or Anthropic’s API sends it to a company your client never agreed to.

Most client contracts list which outside companies are allowed to handle their data, and a model vendor is not on that list.

Adding one means telling every client and usually asking them to sign off, while the data sits somewhere you did not choose for as long as someone else decides. Running the model yourself removes the question instead of answering it.

The platform exposes itself over MCP, so an agent can work against your real findings, fields and templates instead of guessing at them.

Built for agents as well as people

PentestPad is built to be driven by software as well as by a person at a keyboard, and the model doing the driving can sit inside your own network.

That is the whole answer to the problem above: the assistant is pointed at a self-hosted LLM, and client content never crosses your perimeter.

The writing assistant works from what a tester already has. Hand it scanner output, a Burp finding, a screenshot or a log file, attach the evidence, and it returns a structured finding written to the instructions your organisation has defined, so the draft arrives in your format rather than a generic one.

Underneath, PentestPad exposes itself over MCP. Methodology, past findings, templates, finding fields and the rules for writing them live in the platform, and the MCP server exposes fourteen typed tools over them, from searching projects and findings to filing a new one from a template, updating remediation after a retest, or adding a comment.

The platform stays the source of truth and the agent holds no state of its own.

Run end to end the stack is three parts, all on hardware you control. PentestPad is the knowledge base, a local model does the reasoning, and a terminal harness is wired to both.

PentestPad’s own team uses Qwen3 Coder through Ollama, with OpenCode or Claude Code as the harness.

For teams that would rather not assemble one, there is a command-line agent for macOS, Linux and Windows, carrying six customisable specialists across web, network, reconnaissance, API and mobile testing, plus a general one for methodology and tool selection.

How the report gets built

A project holds the client, the assessment type, the scope, the testers assigned to it and the timeline.

Findings attach to it as testing progresses, so what is in scope, who is working on it and what has been found so far sit together rather than across a folder, a chat thread and a spreadsheet.

Each finding carries a title and description, the affected assets, severity, remediation, references and the supporting evidence.

Severity is set with a built-in CVSS calculator, and where a team hides its own impact and probability fields the risk rating derives from the CVSS score instead.

Findings do not have to start empty. A library of reusable vulnerability templates sits alongside the projects, searchable by name and filterable by assessment domain, and importing one fills every field of a new finding with your own house description, remediation text and CVSS vector, leaving only the engagement-specific detail to write. Executive summaries work the same way.

Findings move through statuses your team defines, from draft to ready. Comments attach to individual vulnerabilities rather than to the document, and revision history shows what changed between drafts.

The report itself is built from your own file, with placeholders dropped in where the data belongs, and the PentestPad team does the conversion at no cost. For teams starting from scratch there is a free penetration test report template to work from.

Generation pulls the structured data through it. Tables of contents regenerate on every build, charts and tables populate from the findings, and cross-references resolve through a dynamic link system so they stay correct when content moves.

Reusable field templates cover report-level and finding fields, and a live preview renders the document in the browser before anything is exported.

What comes out covers executive summary, scope, methodology, findings, technical detail, remediation recommendations and supporting information.

A single project can span more than one assessment domain, with each finding tagged to the domain it came from, so a web application test and a network test for the same client arrive as one report with the findings grouped rather than as two separate deliverables.

A library of template functions ships with the platform, from parsing a CVSS vector and deriving a risk rating to counting findings by severity or filtering scope down to just the IPs or just the URLs.

You can define your own JavaScript functions and call them from the template. Reports generate over the API, so a pipeline can produce one without a human opening the platform, and a finished PDF can be encrypted with a password you supply.

Individual DOCX or PDF files can be swapped inside an existing report without rebuilding it.

Branding, access and retention

Whitelabeling covers the logo, the colours and the domain, so someone logging in to check progress arrives at your brand rather than the vendor’s, down to the 2FA challenge screen, and exported reports carry the same identity.

Access control is granular rather than role-shaped. Pentester, Manager and Admin ship as defaults, but you can define your own roles from a permission set covering projects, findings, reports and clients, assign more than one role to a single user, and override individual permissions on top of a role where it does not quite fit.

Role administration is itself a separate permission, so team administration can be granted without also handing over the ability to edit roles. Client accounts are created and invited over the API, and client API keys are restricted separately from your team’s.

Uploads are scanned for malware and flagged files rejected, though anyone with the permission can override that and upload a known sample anyway. Scanning is always on for the managed service and optional when self-hosted.

The data lifecycle is configurable in the same way. Inactive projects archive automatically, and archived reports can be deleted on a schedule you set rather than accumulating until someone remembers them.

Deployment, compliance and pricing

Self-hosting is available on every tier, which matters when a client contract or an internal policy restricts where report data may live.

Self-hosted instances check a version endpoint and raise update notifications, so an isolated deployment knows when it has fallen behind.

Project data, evidence, the finding library, penetration test report template and user configuration export in DOCX, PDF, XLSX or over the API.

Quotes

Available if useful. Use either, both or neither.

“Where a client’s findings are allowed to live is decided by the team holding them, and for many that means nowhere outside their own infrastructure. That is the requirement we built the AI around, so the model runs wherever a team needs it to, including entirely inside their own network.”

Luka Šikić, CEO, PentestPad, Secure Block.

“Most teams reshape the fields, the layout and the statuses to match how they already work, set which parts a client can see, put their own branding on the portal and write their own instructions for how the AI should fill a finding in. That was the point of building it that way.”

Mirna Novak, Director, PentestPad

The post PentestPad: How Pentest Companies Adapt In The Era of AI appeared first on Cyber Security News.

Cybersecurity Awareness Month 2026 arrives as online attacks become easier to launch, harder to spot, and more focused on people. Observed every October, the campaign helps individuals and businesses build safer digital habits.

This year there are two themes. The NCA is running “Don’t Make It Easy for Them”, a push to make life harder for cybercriminals through small daily habits. CISA is running “Securing the Next 250”, tied to America’s 250th anniversary, with a strong focus on critical services like power and water.

The message is not limited to the U.S. Canada’s Get Cyber Safe program launched its own Cyber Month under the theme “Your best defence is you”, warning that AI is making scams look more real than ever.

After years of covering breaches, one pattern stands out to me: most incidents still start with a person being tricked, not a system being cracked. That is why this year’s focus on habits matters so much.

Phishing is no longer limited to badly written emails. Attackers now use fake CAPTCHA pages, QR codes, Teams chats, AI-written messages, OAuth flows, voice calls, and real-looking sign-in pages.

Microsoft said in its Q1 2026 email threat report that it detected about 8.3 billion email-based phishing threats during the first quarter of 2026. By the end of the quarter, QR-code phishing had become its fastest-growing attack vector.

Cybersecurity awareness is not a once-a-year training task. It is an everyday defense layer.

Why Cybersecurity Awareness Month Matters in 2026

Modern attacks often start with a normal action: opening an email, scanning a QR code, approving a sign-in, or following a “security verification” step. The goal is to make the request feel routine.

CyberSecurityNews has covered several clear examples. Recent ClickFix fake CAPTCHA attacks have convinced users to paste malicious commands into Windows Run. Other attackers have impersonated IT help-desk staff on Microsoft Teams and asked employees to install software or provide remote access. CyberSecurityNews has also reported on OAuth device-code phishing attacks targeting Microsoft 365 accounts.

These cases show why people still matter. Security tools stop many threats, but users must spot when a normal-looking action does not make sense.

10 Important Cybersecurity Habits to Stay Protected Every Day

1. Use a Unique Password for Every Account

Never reuse the same password across email, banking, social media, work accounts, and cloud services. If one site is breached, reused passwords can let attackers try the same login elsewhere.

2. Use a Password Manager

A trusted password manager can create and store long, random passwords. CISA includes password managers in its core online safety advice.

3. Turn on MFA Wherever Possible

Enable multi-factor authentication for email, cloud, banking, developer, and admin accounts. Prefer passkeys, hardware keys, or authenticator apps where supported. Never approve an MFA request you did not start.

MFA remains important, but users should also understand that modern phishing kits can steal authenticated sessions after MFA has been completed. CyberSecurityNews recently documented attacks where victims completed MFA but attackers captured the Microsoft 365 session.

4. Keep Devices and Apps Updated

Install security updates for Windows, macOS, Linux, Android, iOS, browsers, VPNs, routers, and business software. Attackers often scan for systems that have not received known security fixes.

5. Stop Before Clicking

Urgency is a common social engineering tool. Messages saying “act now,” “account locked,” “payment failed,” or “CEO needs this today” should make you slow down and verify the request through another channel.

6. Check the Real Destination

Before entering a password, inspect the site name. Attackers use look-alike domains, redirects, fake login pages, and trusted cloud services to hide the final destination.

Do not assume a page is safe simply because it contains Microsoft, Google, DocuSign, Cloudflare, or another known brand.

7. Treat QR Codes Like Links

Do not scan an unexpected QR code because it appears on a trusted-looking email, poster, invoice, or document.

Unit 42 researchers reported detecting about 75,000 QR codes each day, with around 15% of the pages containing QR codes leading to malicious links. CyberSecurityNews has also detailed how QR codes are being used for phishing and malicious mobile app delivery.

8. Never Run Commands From a Website

A web page should not normally tell you to press Win+R, open PowerShell, launch Terminal, or paste a command to “verify” yourself.

This is a key warning sign of ClickFix. Microsoft has documented ClickFix campaigns where fake CAPTCHA, download, and verification pages persuaded users to run attacker-provided commands.

9. Back Up Important Data

Keep important files backed up using a method that ransomware cannot easily change. For business systems, test recovery regularly. A backup matters only if it can be restored.

10. Report Suspicious Activity Quickly

Report phishing emails, unusual MFA prompts, fake IT messages, and unexpected login alerts. Fast reporting helps security teams block domains, revoke sessions, reset accounts, and protect other users.

10 Social Engineering Attacks to Watch in 2026

1. AI-Written Phishing

Generative AI helps attackers create clean, personal phishing messages. Grammar mistakes are no longer a useful warning sign. Focus on the request, sender, link, and context.

CyberSecurityNews has detailed on how AI-generated phishing can steal authenticated browser sessions without traditional malware.

2. Adversary-in-the-Middle Phishing

Adversary-in-the-middle, or AiTM, phishing places attacker infrastructure between the victim and the real login service.

The victim may enter the correct password and MFA code, but the attacker captures the authenticated session and can use it to access the account.

3. OAuth Device-Code Phishing

This method sends victims to a real Microsoft sign-in page and asks them to enter an attacker-provided device code.

The website itself may be genuine, but approving the code can give the attacker access tokens. Proofpoint reported rapid growth in device-code phishing during 2026 as public tools and phishing services became easier to obtain.

4. ClickFix and Fake CAPTCHA Attacks

ClickFix pages tell users to run a command to complete a CAPTCHA, fix a browser issue, install an update, or open a file. In 2026, they targeted both Windows and macOS users.

The danger is that the victim performs the execution step for the attacker.

5. QR-Code Phishing

Also called quishing, these attacks place malicious links inside QR codes. The victim often moves from a protected desktop email environment to a mobile browser, where the final page may steal passwords, MFA codes, payment information, or other sensitive data.

6. Fake IT Help-Desk Messages

Attackers impersonate support staff through Teams, email, chat, or phone calls. They may ask the victim to open Quick Assist, install remote-control software, share a code, reset a password, or give them control of a computer.

7. Voice Phishing and AI Voice Cloning

A caller may pretend to be a manager, bank worker, supplier, or family member. AI voice tools can make the call more believable.

Verify sensitive requests through a known phone number or another trusted channel rather than trusting the incoming call.

8. Business Email Compromise

Business email compromise, or BEC, attackers impersonate executives, suppliers, finance teams, or customers to request payments, invoice changes, gift cards, or sensitive files.

A compromised real mailbox makes these messages especially convincing because the attacker may have access to previous email conversations.

9. Fake Job and Recruitment Attacks

Attackers use fake job offers, interview tasks, coding tests, and recruiter messages to steal login details, personal data, cryptocurrency, or developer credentials.

Some campaigns also deliver malware through project files or software packages.

10. Consent and App-Permission Scams

Instead of stealing a password, attackers may ask a user to approve a cloud app, OAuth connection, browser extension, or account permission.

One careless approval can provide long-term access to email, files, contacts, customer records, or other business data.

What Businesses Should Do During Cybersecurity Awareness Month

Organizations should use October to strengthen real habits, not just send yearly training slides. Run phishing drills, review MFA, remove unused accounts, test backups, patch exposed systems, and make reporting easy.

Security teams should also train staff on newer attack paths. Employees need to understand that a genuine Microsoft login page can still be part of phishing, a CAPTCHA can be malicious, and a Teams message from “IT Support” may actually come from an external attacker.

Good awareness programs avoid blame. People report mistakes faster when the process is simple, giving defenders time to reset passwords, revoke tokens, isolate devices, and stop a small event from becoming a larger breach.

Cybersecurity Awareness Month 2026 is a reminder that strong security is built from small actions repeated every day. The theme “Don’t Make It Easy for Them” fits the current threat landscape well.

Use unique passwords, enable MFA, update software, question unexpected requests, check links, avoid copied commands, and report suspicious activity early. Attackers will keep changing their tools, but these habits remove many easy paths they depend on.

Cybersecurity does not require every person to become a security expert. It requires people to recognize risky moments, slow down, verify unusual requests, and make the safer choice.

The post Cybersecurity Awareness Month 2026 – Don’t Make It Easy for Them appeared first on Cyber Security News.

Sony has moved to protect PlayStation 5 consoles with its latest system software update, as the newly released Relapse jailbreak draws attention to a broad set of older PS5 firmware versions.

The update matters for console owners because the public Relapse exploit supports PS5 and PS5 Pro systems running firmware versions 7.00 through 13.60, allowing unsigned code to run on compatible devices.

Sony released its firmware 14.00 update before Relapse became public, rather than directly after. However, the timing has made the difference clear: PS5 consoles that have installed firmware 14.00 are outside Relapse’s stated supported range.

PS5 Security Update

Sony has not publicly described specific Relapse fixes on its consumer update page, but it says PS5 system software updates include security updates needed for the console to work properly and strongly recommends installing them quickly.

Relapse was published as an exploit chain for PS5 firmware 7.00 to 13.60. It works on both the original PS5 and PS5 Pro models within that version range.

According to the project’s GitHub documentation, the jailbreak first targets the console browser through a WebKit weakness, then uses a kernel-level flaw to gain read and write access to the system. This level of access can allow compatible homebrew tools and other unsigned payloads to run on the console.

The exploit is not a permanent jailbreak. Relapse is tethered, meaning users must run the exploit again after every PS5 restart. Its developers also warn that the browser stage may require several attempts and that the kernel stage can freeze or crash the console.

The project itself lists risks including system instability, data loss, and PlayStation Network account bans. It also states that the software is intended only for education and security research, and does not support piracy or unauthorized use.

For everyday PS5 owners, updating remains the safer path. A console on the latest supported system software receives Sony’s current security protections, feature changes, and better compatibility with games and PlayStation services.

Sony warns that users may be unable to use some apps or content without updating system software. It also cautions against installing update files from unofficial sources or on altered consoles, as this can cause the system to stop working correctly and may affect warranty or repair support.

PS5 users can check for an update by opening Settings, selecting System, then choosing System Software and System Software Update and Settings. If an update is available, users can select “Update System Software” to begin installation.

Sony also supports automatic downloads and USB-based updates for consoles that cannot complete a normal download. For USB updates, the company requires a FAT32 or exFAT drive with a PS5/UPDATE folder structure and the official update file saved as PS5UPDATE.PUP.

The Relapse release shows why console firmware remains a key security boundary. A browser flaw alone is serious, but combining it with kernel access can weaken the controls designed to prevent unapproved code from running.

Firmware 14.00 creates a practical dividing line for now: systems on 7.00 through 13.60 may be exposed to the Relapse chain, while consoles already updated to 14.00 do not fall within the jailbreak’s published compatibility range.

Since Sony does not provide an official way to downgrade PS5 system software, users who update cannot simply roll back to a vulnerable release.

Cut every SOC alert investigation by 21 min. Power your SOC with instant IOC context for immediate response: Integrate TI Lookup into your SOC

The post Sony Rolls Out PS5 Security Update Following the Release of Relapse Jailbreak appeared first on Cyber Security News.

WordPress administrators are being urged to update after researchers disclosed Click2Shell, an exploit chain that can turn one malicious link into remote code execution (RCE) on a vulnerable website.

The issue begins with a WordPress Core theme-preview weakness that silently installs an attacker-selected theme from the official directory, then becomes a server compromise when chained with insecure pre-activation code in a theme.

WordPress addressed the Core flaw in version 7.1.1, released September 17, 2026, as part of an update containing 11 security fixes, 17 Core bug fixes, and 19 Block Editor fixes.

The official advisory describes the issue as specially crafted URLs automatically installing and previewing an inactive WordPress.org theme, credits Paulos Yibelo and pwn.ai, and recommends immediate updates.

The attack does not require the adversary to hold a WordPress account. It does, however, require a logged-in administrator to visit the malicious URL.

Click2Shell WordPress Vulnerability

The administrator’s authenticated session provides the installation capability and nonce, allowing trusted WordPress JavaScript to perform the sensitive action on the attacker’s behalf.

The vulnerability originates in the theme installer route, where a theme value is processed in two inconsistent ways. WordPress.org’s Themes API canonicalizes the supplied input into a valid catalog slug, while the administrator’s browser retains the original punctuation and inserts it into a jQuery selector.

Crafted selector characters can escape the intended attribute match, traverse the returned theme card, and reach the genuine Install control, which WordPress then activates programmatically.

This behavior alone does not permit an arbitrary theme archive or immediate RCE. It installs a current package selected from the trusted WordPress.org catalog, and the theme remains inactive, leaving the site’s appearance unchanged. That lack of visible disruption makes the activity easy to miss and explains why the second stage is crucial.

Researchers demonstrated the complete chain with Mobile Repair Zone 2.5.4. During a Customizer preview, WordPress loaded the inactive theme’s PHP, exposing an authenticated AJAX handler that lacked both a nonce check and a capability check.

The handler accepted attacker-controlled plugin details and a package URL, downloaded and unpacked the supplied archive, and loaded its PHP entry point, producing code execution under the web server’s account.

A successful compromise could allow attackers to access wp-config.php and database credentials, read WordPress or WooCommerce data, alter files and content, create users, steal secrets available to the PHP worker, and potentially seize the wider hosting environment.

The Core weakness therefore acts as a forced-install primitive, while the vulnerable theme supplies the bridge from an inactive package to executable attacker-controlled PHP.

WordPress fixed the selector injection in changeset 63664. The patch restricts matching to a genuine div.theme card and applies jQuery’s escapeSelector() to the URL-derived slug before constructing the selector. As a result, injected quotation marks, combinators and comment syntax are interpreted as literal slug characters instead of executable CSS selector structure.core.trac.

Pwn.ai assessed the standalone forced-install issue as High severity with a CVSS 3.1 score of 7.1 and considered the demonstrated RCE chain Critical. WordPress had not published a final severity or CVE identifier at disclosure, and public reporting indicated no evidence of exploitation in the wild.

The disclosure process began on August 22, when researchers reported the selector injection and automatic installation behavior. They delivered the full pre-activation chain on September 1, WordPress shipped its correction on September 17, and pwn.ai published the detailed technical analysis the following day.

Site owners should install WordPress 7.1.1 immediately or apply the corresponding security release for their supported branch. WordPress said fixes were being backported where necessary to security-supported branches through 4.7, although only the latest version is actively supported.

Administrators should also verify automatic updates, review recently installed themes and plugins, inspect unexpected PHP files and account changes, and investigate suspicious requests to theme-install.php or Customizer-enabled admin-ajax.php endpoints.

Cut every SOC alert investigation by 21 min. Power your SOC with instant IOC context for immediate response: Integrate TI Lookup in your SOC

The post Click2Shell WordPress Flaw Lets Attackers Gain RCE With a Single Malicious Link appeared first on Cyber Security News.