Malwarebytes and Claude integrate to scan suspicious messages

Verifying messages with AI

Malwarebytes has launched a new connector for Claude that helps users identify potential scams in their communications. This tool allows people to check suspicious links, phone numbers, and email addresses directly within the Claude chat interface. Instead of navigating to a different website, users can simply provide the information to the AI for analysis.

The integration is designed to be accessible to a wide audience by working with both free and paid Claude plans. One of the key features is that users do not need to create or maintain a separate Malwarebytes account to use the connector. This lowers the barrier for people who want to verify messages but do not want to manage extra security software.

According to Help Net Security, the connector can process various types of input. Users can paste a specific link into the chat or upload a screenshot of a message they received. The AI then uses these inputs to cross-reference details against the Malwarebytes threat database to determine if the content is safe.

How the connector tools work

The connector provides six distinct tools that Claude can use to perform different types of security checks. These include the ability to look up specific domain registration details and the option to scan several items at once. It also includes a specific tool for reporting suspected scams to the relevant authorities.

Users have granular control over how these tools interact with the AI. You can set permissions for each of the six tools separately to decide how the system behaves. There are three primary settings: 'Ask' requires Claude to request permission before every use, 'Always allow' lets the tool run without prompting, and 'Block' hides the tool entirely.

When a user chooses the 'Ask' setting, Claude will pause and request permission before accessing any specific data. This ensures that the AI only uses the security tools when the user explicitly gives the go-ahead. This permission model helps users maintain control over their data while still benefiting from the automated scanning capabilities.

Real-world scam detection

In a practical test of the connector, a screenshot of a text message from a supposed government agency was analyzed. The message claimed the recipient had an unpaid speeding fine and warned that it was a third notice. The AI requested permission to check the sender's phone number, which Malwarebytes identified as 'Unknown.'

Claude then flagged the message as a likely scam based on several red flags. It pointed out that the sender used a Philippine country code despite claiming to be from Croatia. It also noted the high-pressure language used in the text to demand immediate action from the recipient. This helped clarify why the message was deceptive without the user having to manually research the details.

Another test involved two delivery updates from a major furniture retailer that contained shortened links. Both links were marked as 'Unknown' by the scan, but the AI was able to see that one link redirected toward an order page. The other link was marked as 'Offline.' By seeing the redirect, the user could verify the information in a separate app without actually clicking the link in the text.

Understanding the risks

The primary goal of the Malwarebytes and Claude integration is to help users understand risks before they interact with a message. By providing context on why a message looks suspicious, the tool moves beyond a simple 'safe' or 'unsafe' label. It explains the specific details, such as mismatched country codes or suspicious redirects, that indicate a scam.

This approach is particularly useful for identifying sophisticated phishing attempts that might not be caught by standard filters. Because the AI can interpret the content of a screenshot or a long text, it can catch nuances in the language. It helps the user see the pattern of the scam rather than just the technical data of the link.

The integration serves as a bridge between a massive threat database and a conversational interface. It allows the user to ask questions about a suspicious message and receive an explanation of the risks. This helps people make informed decisions about whether to reply to a text or click on a link provided in an email.