Craig M.

Craig M.

London, United Kingdom · Chief Product Officer @ Cyberr®

Access Creative College

About

Craig M. is a technical founder and VP of Product who builds AI-driven systems that eliminate manual work across revenue, hiring, product, and operations. His focus is on taking AI from concept to production by designing scalable workflows, custom APIs, and LLM-powered automation that deliver measurable business outcomes.

-- --

-----

-

Specialties

  • Ethical Hacking
  • Hardware Hacking
  • Network Penetration Testing
  • Penetration Testing
  • Social Engineering

Skills

  • Large Language Models (LLM)
  • Artificial Intelligence (AI)
  • Product Management
  • Go-to-Market Strategy
  • Sales Operations
  • Claude Code
  • Lovable (Diamond Builder)
  • Automation Systems
  • Clay
  • APIs

34 skills in total

Experience

  • Chief Product Officer, Cyberr®

    CYBERR

    Jan 2026 — Present

    London Area, United Kingdom

  • Founder & CEO

    Recruiterbolt

    Apr 2021 — Dec 2025

    United States / Remote

  • VP of Sales

    Breadwinner

    Jan 2022 — Nov 2022

    London, United Kingdom

10 experiences in total

Education

  • Access Creative College

    Bachelor's degree, Popular Contemporary Music

    Sep 2007 — Sep 2009

Recent posts

3/3 on today's Cyberr daily challenge.

Senior level: vulnerability management, threat intelligence and application security.

Can you match 3/3 on Vulnerability management?

https://cyberr.ai/daily/8e8811f8-c29e-4d35-aaa1-8f82ec12a2c4

View post

The most dangerous "deepfake" in the FBI's new North Korea alert mostly isn't a deepfake.

On 12 November the FBI and DOJ detailed FaceStealer: more than $400M in crypto stolen over 18 months, and over 900 victim webcam videos seized.

The chain is simple, which is why it works:

→ Hijack a real person's Telegram, LinkedIn, Slack or Discord.

→ Invite their contacts to a "catch up" on a look-alike Zoom or Teams domain.

→ Play that person's genuine, stolen webcam footage.

→ Claim the target can't be heard, then send a "fix" to run in the terminal.

→ Record the new victim's webcam. Repeat.

*The FBI says the scheme still relies overwhelmingly on real victim video, not AI.*

So "spot the deepfake" training won't save you. The face is real. The account is real. Only the request is fake.

It's also moving upstream. The FBI describes a near-miss against the maintainer of a JavaScript library with tens of millions of weekly downloads, plus attempts on npm and Rust crate maintainers.

*Your dependency's maintainer's laptop is part of your attack surface.*

What I'd check this week:

→ One rule, said loudly: no legitimate meeting ever needs you to run a script or paste a command. Teach the "audio fix" pattern by name.

→ Join only via the native client or your managed meeting domains, and filter newly registered look-alike domains.

→ Alert on terminal or Run-dialog commands that fetch and execute remote code (curl piped to sh, PowerShell download cradles, osascript).

→ Pin dependencies and add a cooldown before adopting new package versions, so a poisoned release has time to be caught.

Verify the action, not the face.

Who has blocked paste-to-terminal on developer endpoints, and how much friction did it cause?

#ThreatIntel #SupplyChainSecurity #SocialEngineering

View post

Your ScreenConnect console had MFA. The extension didn't ask for it.

On 11 November, Huntress saw rogue ScreenConnect clients and a toolkit called SCKit pushed to at least 288 endpoints across 48 organisations in about 90 minutes. A second wave hit 205 hosts on 13 November.

The likely door, per Huntress (ConnectWise says initial access is still unconfirmed): RMM+, a third-party extension that lets another tool drive ScreenConnect sessions using a shared access key. That path sidesteps ScreenConnect accounts, roles and MFA. In versions 1.0.31 and 1.0.32, if nobody had ever set the key, a blank one was accepted. Most installs were legitimate. Added for a project, then forgotten.

The nastier part: SCKit is built to compromise the technician who connects to an infected machine to investigate, then reuse their session and permissions. Huntress hasn't seen that spread succeed. The intent is still clear.

What I'd check this week if you run ScreenConnect or any RMM:

→ Inventory every extension on every instance, on-prem included. ConnectWise disabled the suspect ones on hosted instances; on-prem is on you.

→ Treat every shared key or API path as an identity: owner, where it's set, last rotation.

→ Confirm technician clients run 26.6.5 or later, and audit Run Command and Toolbox history.

→ Investigate suspect hosts out of band (EDR, WinRM), never through the RMM session itself.

Continuous control checks matter here. In Heelr by Cyberr, Heelr Compliance automatically checks things like whether every Microsoft 365 user has MFA enforced. But any check only covers the doors you know exist.

When did you last list every way into your RMM that isn't a named user with MFA?

#MSP #RemoteAccess #IdentitySecurity

View post

Fireworks AI was breached twice in three weeks. The second time, the way in was someone else's cloud.

From the company's own update (9 Nov): in early October, an intruder used internal credentials to capture secrets from evaluation jobs. Fireworks rotated keys and tightened access.

Then on 23 October, a third-party cloud provider hosting Fireworks' R&D machines had its own incident. The next day, the attacker used that foothold to reach Fireworks development machines and run a workload that exposed billing data and, for some customers, datasets and incomplete model weights.

Fireworks links the actor to a financially motivated "LLMJacking" group: steal access, resell model tokens. That is a recurring revenue motive. Rotating a key removes the key. It does not remove the attacker's reason to come back.

My takeaway: your blast radius is set by your least-governed compute, and R&D is usually it.

What I'd check this week:

→ Inventory every place dev and research workloads run, including GPU and niche cloud providers outside your main tenant.

→ Map what those machines can reach. If a dev box can read customer datasets, it is production.

→ Treat a hosting provider's security notice as your own incident: rotate credentials and tokens that live on that infrastructure the same day.

→ After "containment", keep hunting for the same actor through different doors for at least 30 days.

Honest question for security leaders: who in your org owns the security of the GPU box someone rented last quarter?

#CloudSecurity #ThirdPartyRisk #AISecurity

View post

3/3 on today's Cyberr daily challenge.

Senior level: detection engineering, threat intelligence and application security.

Can you match 3/3 on Detection engineering?

https://cyberr.ai/daily/32a52c24-067c-48cb-8fc5-5b4805ca3d47

View post

View Craig M.'s profile on Cyberr