You know that feeling? You're developing a new email feature, you run your test script, and boom you realize 3 seconds too late that you used the production database. Your CEO just received an email with the subject TEST - DO NOT READ - LOREM IPSUM. Or worse: you configured a cloud SMTP server for testing, forgot to disable actual sending, and now your Mailgun account is suspended for suspicious activity because you sent 847 emails to test@example.com in 5 minutes.
An FBI informant helped run the Incognito dark web market and allegedly approved the sale of fentanyl-laced pills, including those from a dealer linked to a confirmed death, WIRED reported this week. Meanwhile, Jeffrey Epstein's ties to Customs and Border Protection officers sparked a Department of Justice probe. Documents say that CBP officers in the US Virgin Islands were still friendly with Epstein years after his 2008 conviction, illustrating the infamous sex offender's tactics for cultivating allies.
For Gmail users, there is an automatic opt-in that may allow Google access to your emailed data (think: your personal and work messages, your attachments) "to train AI models," cybersecurity experts allege. If you don't want this information shared, you need to adjust your settings. In the race for companies to get an ROI on AI, we're already seeing language learning models running out of new, human-generated data to train on.
The scheme involves a text message that appears to come from 'Gmail from Google,' warning recipients that their account has been compromised. The message includes a link labeled 'Recover Account.' When users click it, they are prompted to enter their Gmail password, which is then captured by scammers. In some cases, attackers can combine stolen information with personal details, like your phone number.
Gmail's version lets you ask questions about your messages in the search bar, using natural language. Google uses the example of, "Who was the plumber that gave me a quote for the bathroom renovation last year?" It's hard to imagine that saving much time over a basic search for "plumber quote" or "plumbing estimate," but maybe it could help in some situations.
Web browsers are among the top targets for today's cybercriminals, playing a role in nearly half of all security incidents, new research reveals. According to Palo Alto Networks' 2026 Global Incident Response report, an analysis of 750 major cyber incidents recorded last year across 50 countries found that, in total, 48% of cybercrime events involved browser activity. Individuals trying to connect to the web, including business employees, are exposed to cyberthreats on a daily basis.
This Privacy Notice applies to all personal information processed by CoinDesk, including its affiliates and subsidiaries (" CoinDesk," " we," " us," or " our "). It covers the information CoinDesk collects through the websites, mobile applications, electronic devices, all other products and services we provide, any other services that display this Privacy Notice, all of the associated content, functionalities, and advertising, and when you communicate with us by phone, email, or otherwise (collectively, the " Services ").
While you're thinking about third-party add-ons for your computer and phone, take a moment to review everything you have installed on both fronts and consider how many of those programs you actually still use. The fewer cracked windows you allow on your Google account, the better - and if you aren't even using something, there's no reason to keep it connected.
Meanwhile, the actual threat landscape evolved in an entirely different direction. Today's attackers aren't sitting at keyboards manually typing password guesses. They're running offline brute force attacks with dedicated GPU rigs that can attempt 100 billion passwords per second against hashing algorithms like MD5 or SHA-1. At that speed, your clever substitution of "@" for "a" buys you microseconds of additional security.
This is because these unsubscribe links usually take you to a web page via a URL embedded in the unsubscribe text that identifies your email address, either in plain text or via an alphanumeric code. The moment this unique URL loads, the spammer at the other end knows that you were the one to click it; they now know that the email address they blasted does, in fact, have a real person at the other end.
In a service alert spotted by BleepingComputer, Microsoft revealed that the glitch started on February 5 and has been preventing some Exchange Online users from sending and receiving emails. "Some users' legitimate email messages are being marked as phish and quarantined in Exchange Online," Microsoft said in the service alert. "We've determined that the URLs associated with these email messages are incorrectly marked as phish and quarantined in Exchange Online due to ever-evolving criteria aimed at identifying suspicious email messages, as spam and phishing techniques have become more sophisticated in avoiding detection."
Earlier this month, Joseph Thacker's neighbor mentioned to him that she'd preordered a couple of stuffed dinosaur toys for her children. She'd chosen the toys, called Bondus, because they offered an AI chat feature that lets children talk to the toy like a kind of machine-learning-enabled imaginary friend. But she knew Thacker, a security researcher, had done work on AI risks for kids, and she was curious about his thoughts.