Study finds ChatGPT, Gemini and Perplexity sometimes send shoppers to fake retail sites
IP protection firm Podqi asked the chatbots shopping questions about 151 brands and found they could point users to imposter and other risky websites. The test ran over two periods starting in September.
If you use chatbots to find where to buy things, double-check the website address before entering payment details.
aii.news picks and summarizes stories from trusted news sources with the help of AI. Every story links to the original reporting. How we work.
Keep reading
-
186 lookalike AI tool domains push ClickFix lures, browser malware and search hijackers
Researchers at Island found lookalike domains posing as popular AI tools that trick visitors into fake verification steps, install browser-hijacking malware or add extensions that hijack searches. The technique has been named TypoPrompting.
Why it mattersTyping an AI tool's address wrong can lead to malware, so bookmark the real sites and be wary of any page asking you to run commands.
-
United Airlines blocks customers from using AI agents to book and manage flights
United says it will not let personal AI assistants book tickets or manage reservations, warning that sharing account credentials with third-party agents can expose sensitive data and cause unintended purchases. It also cited federal rules on information airlines must give passengers.
Why it mattersCompanies are starting to restrict AI agents on their accounts, which may limit what consumer agents can do for you online.
-
Google releases Gemini 4 Hydrogen, now the default model in Search AI Mode
Google launched Gemini 4 Hydrogen, which it says handles multi-part questions, typos and very specific requests better, and it replaces the Flash series. It is available today in Search, the Gemini app, AI Studio, the Gemini API and Gemini for business.
Why it mattersBillions of searches and many developer apps will now run on this model, and independent tests suggest it offers near-top quality for less money than some rivals.
-
Chinese-speaking hackers used open-source AI pentest tool to breach 10 organizations
Palo Alto Networks Unit 42 says attackers paired the open-source AI pentesting framework Cairn with Chinese models GLM and DeepSeek to run 25 operations against at least 19 organizations in August and September 2026. They gained access or took data at 10 targets, mostly in semiconductors, advanced manufacturing and defense, often after giving little more than a domain name and a one-line goal.
Why it mattersAutonomous AI attack tools are now being used in real campaigns, so defenders must assume attacks can run at machine speed with little human effort.
-
Google releases Gemini 4 Carbon, its strongest model yet, with an agent 'swarm' feature
Google is rolling out Gemini 4 Carbon to the general public after first giving it to a select group of cybersecurity defenders. Google says it pushes beyond its Argon model, though Argon remains the suggested choice for most coding and knowledge work.
Why it mattersDevelopers and users now have a new top-tier Google model to compare against the best systems from Anthropic and OpenAI.
