Welcome to The Senroc Download! This is your weekly briefing from the team at Senroc Technologies letting you know what we're seeing, what we're thinking, and tips you can use right now.

An AI research non-profit called METR published an account of what happened after one of its API keys was left exposed. A dashboard that was supposed to sit behind a login silently stopped requiring a login, attackers found it, they asked the AI agent to reveal its key, and then spent roughly $600,000 of API credits over three weeks. The key wasn’t cracked. It was simply reachable, and handed over on request.

The same week, Anthropic logged users out of Claude accounts and stripped saved payment methods after finding that ordinary credential-stealing malware — the commodity kind that scrapes everything on an infected PC — was harvesting AI session cookies. Those cookies allow bad actors right past multi-factor authentication because as far as the service is concerned the sign-in already happened and was approved.

What both of these events have in common is that nobody set out to attack AI. AI accounts have simply joined the list of things worth taking, alongside bank logins and email. They’re attractive for a specific reason: usage costs real money, the bill arrives later, and almost nobody is watching the meter while it’s climbing.

Takeaway: Two practical items. If your business pays for AI by usage rather than a flat fee, put a spending alert on it today, so a stolen key shows up as an alert rather than an invoice. Secondly, treat AI logins the way you treat your bank account: unique password, multi-factor authentication, and no single account shared around the team. If anyone has ever pasted an API key into a script, a spreadsheet, or a chat message, that key should be replaced immediately.

Security firm Huntress documented a campaign that opens with an ordinary phishing email dressed up as an invoice. The link leads to a page that hands the victim a genuine, properly signed installer for Faronics Deploy — a legitimate tool IT departments use to manage computers remotely — disguised as an Adobe update. Running it quietly enrolls the machine into the attacker’s controls.

From there the attacker uses the tool exactly as an administrator would: run a script, install additional software, and install ScreenConnect, a real remote-support product. The campaign ran from late July into August and reached more than 450 computers before the vendor stepped in stop it.

The reason this works is that none of what is installed is actually malware. Every component is a real product with a valid signature, doing precisely what it was built to do. Security software looking for malicious files ignore all of it because they are plausible applications.

Takeaway: Two critical things here. Can you answer what remote access tool, or tools, your IT provider uses? If not, get the answer in writing so you can be aware — most providers use one, occasionally two. Anything else that turns up needs to be addressed immediately. Second, this threat doesn’t work unless a human makes a mistake by falling for the phishing email. Educate your employees about these kinds of threats so they can be on the lookout, and if they’re not doing regular cybersecurity training, they should be!

When a business email account is broken into, one of the first things an attacker sets up is a rule that forwards a copy of everything — or everything containing words like invoice or payment — to an outside address. Then they leave. Nothing looks wrong, mail arrives normally, your password still works, and your account quietly copies everything to someone else for months.

The rule is easy to miss because Outlook offers several places to hide it. Everyone is aware of the normal forwarding on the account itself, but few look in other places: an Inbox rule, a desktop app only rule, or a rule that redirect rather than forward.

The same forwarding feature can be an issue even when it was unintended. An employee forwards their work mail to a personal address so they can keep an eye on things over a holiday, but the forwarding rule never gets removed. Years of company correspondence quietly accumulates in a personal account your business does not control.

Takeaway: Ask your IT provider to block all forwarding to external addresses — it’s a standard setting in both Microsoft 365 and Google Workspace. Then have them run a report of existing forwarding rules across all accounts. It takes only a few minutes! Individually, employees can check their own rules in Outlook under Settings, then Mail, then Forwarding and Rules.