KYC caller names arrive, an APK factory gets busted, and ChatGPT ads get personal

The Department of Telecommunications has told every Indian telecom operator to switch on caller name display, called CNAP, across the country from November 1, 2026, Deccan Herald reports. Incoming calls will show the name on the caller's KYC records, free, and business callers have to show their GST or trademark name. It's the right fix, arriving about four years after TRAI first floated it in 2022, years we spent trusting a crowd-sourced app to tell us who was calling. It won't end scam calls, because plenty of fraud runs on SIMs with someone else's real name attached, but it makes 'I'm calling from your bank' a lot harder to say with a straight face.

Bank of America added about 1,000 contractors in the second quarter just to patch software flaws that AI models, Anthropic's Mythos among them, found in its own and its vendors' code, CEO Brian Moynihan told CNBC. It's the most concrete picture yet of what AI bug-hunting costs the people on defence. Finding the bug is the cheap part now. Fixing it, testing the fix and shipping it across a giant bank's systems and every vendor it depends on is still slow, human work, and apparently a thousand extra humans' worth of it.

If you use ChatGPT on the Free or Go plan, open Settings, then Ad Controls, and switch off Personalize ads, per OpenAI's help page. From then on, ads are picked only from the chat you're in, not your history. Switch off Past chats and memory while you're there, then tap Delete ads data. It takes two minutes, and that three-month-old conversation about your loan EMIs stops deciding what you get sold.

Mumbai Crime Branch has arrested an Indore graphic designer accused of building and supplying 2,805 malicious APK files to fraud networks that took about Rs 150 crore from victims across India, Deccan Herald reported on October 15. One person, thousands of fake apps. That tells you the 'bank update' or 'wedding card' APK in your family group isn't some clever one-off, it's a product line. The myth that keeps it working is that a file from a known contact must be safe, when the contact's phone is usually the first thing that got hacked.

Advertisers can target ChatGPT users by uploading lists of emails, phone numbers and Android ad IDs, which OpenAI matches against accounts, 404 Media found by running its own test campaigns. Gizmodo adds that a list needs at least 25,000 entries. My problem isn't ads in a free product. It's that people tell ChatGPT things they would never type into a search bar, and now the same customer-list machinery that made social media ads feel creepy sits right next to those conversations. OpenAI pitched a different kind of ad business, and this looks like the old one with a nicer chat window.