Black October: How to Protect Yourself From a New Wave of AI-Driven Attacks

AI is making cybercriminals more powerful. Scammers are using it to write more convincing messages, to run attacks at scale, and to find weaknesses in software. At the same time, security researchers using AI tools have discovered serious flaws in websites and platforms that millions of people use daily. Both developments peaked in October 2026, leading the media to dub the month “Black October.”

What is Black October?

“Black October” refers to a wave of AI-driven cyberattacks targeting websites, online services, and organizations around the world in October 2026. It revealed significant vulnerabilities found with the help of advanced AI models. What stands out is the speed: Cybercriminals can use autonomous AI agents to scan databases around the clock and send phishing messages in large volumes.

While the Black October incidents primarily targeted websites and corporate networks, the stolen data can fuel the next wave of phishing attacks. For consumers, this means your online accounts and the personal information they contain may face higher risks.

What happened during Black October?

Several separate and major cybersecurity incidents occurred over the last month:

• Researchers found that hidden instructions in one email could take over an AI assistant: In research released October 1, Salt Security showed that a malicious instruction hidden inside an email could take over Manus, an AI assistant, and reach connected accounts. All the victim had to do was ask the assistant to read their inbox. That flaw has reportedly been patched, so make sure any AI assistant you use is up to date.

• A hacker broke into websites belonging to more than 100 companies: A Reuters investigation published October 19 described how a hacker utilized AI agents to probe more than 1,200 websites, successfully breaching the sites of over 100 companies. The agents did most of the legwork with minimal human input. How much data was actually stolen remains unreported.

• Researchers found significant PHP vulnerabilities: On October 21, researchers revealed serious PHP vulnerabilities. PHP is a popular general-purpose scripting language that powers 70% of all websites in the world, and exploiting it could give criminals control of affected servers and access to steal sensitive information. Fixes exist, but they only help once website operators install them.

• A third party stole credentials from a popular cloud platform: Also on October 21, Railway, a cloud platform developers use to build and host applications, disclosed that unauthorized parties had accessed systems holding customer credentials. Stolen keys and passwords could expose connected services and their data, so Railway urged its customers to update their credentials.

Separately, Axios reported that banks are worried about customers granting AI agents access to their accounts and about those agents possibly doing things they were never told to do. While this is not an example of a cyberattack, it does raise concerns about AI’s increasing access to and control of sensitive accounts.

The hidden risks of AI

Black October showed that AI assistants and autonomous agents are becoming a new avenue for cybersecurity risks. While they can make life easier, connecting them to your email, personal files, or bank accounts gives them significant power and access. If that access is compromised or manipulated, the consequences can include money lost, accounts taken over, and personal data exposed. The Salt Security research is one example of how instructions concealed in content an assistant reads can manipulate it to work against its user.

Avast security experts are currently warning of four distinct risks that consumers should keep in mind over the coming weeks:

• Deceptively realistic follow-up phishing: Scammers are using news of recent data breaches as bait, sending messages like, “Your account was affected by the recent security incident – please verify now” to steal passwords and two-factor authentication codes. Legitimate providers will not ask you to submit login credentials through a link or text message.

• Misleading padlock icons: A padlock or secure connection symbol doesn’t prove a website is safe. Black October incidents involved stolen digital certificate keys, potentially allowing attackers to make fake pages appear legitimate.

• Risks via connected AI tools: Many users have linked third-party tools and add-ons to their accounts. Attackers may try to exploit compromised integrations to gain unauthorized access to email inboxes, cloud storage, or financial services accounts.

• Automated identity theft: Stolen personal information, such as names, phone numbers, dates of birth, and addresses, can give criminals the information they need to commit identity theft. Account alerts can help consumers spot unusual transactions sooner.

Staying safer as AI spreads

To help consumers stay secure, Avast recommends taking basic defensive steps:

• Update passwords: Create new strong passwords for your primary email and financial accounts immediately. Use a unique password for every login, and where available, turn on passkeys or two-factor authentication. A good password manager can help you create and store them.

• Revoke unneeded app permissions: Check your app permissions in your Google, Apple, and Microsoft accounts, and remove any that aren’t essential. Require AI tools and agents to ask for your confirmation before sending messages or making purchases. Never share your bank login with them.

• Avoid clicking suspicious links: Scammers may use news about data breaches to send convincing “your account was affected” messages. Instead of clicking the link in these messages, open the company’s app or type its website address yourself.

• Security software can offer extra protection: Use tools like Avast Free Antivirus, which is designed to offer real-time protection to help detect and block malicious websites and links while you browse, and Avast Assistant, an AI-powered tool that helps you quickly determine whether something is real or fake.

AI is making both attacks and scams more effective, so consumers should take basic precautions now. Reviewing permissions, using unique passwords, and verifying messages independently can greatly reduce the risk to personal accounts.