GoKawiil
Tech News
Microsoft announced it coordinated an industry-wide disruption of EvilTokens, a subscription phishing service sold via Telegram since February that used an AI chatbot to help criminals analyze compromised inboxes and craft convincing fraud emails. The platform compromised 12,000 accounts across 10,000 organizations worldwide, mostly in the US, and Microsoft seized 50 websites and 150 domains as UK police arrested two suspects.
A hobby project called jevchat turns a model referred to as Jev into a chat interface by asking it, one symbol at a time, to predict the next character of its reply given the conversation so far. The tool samples from a probability distribution over a chosen alphabet (letters, truncated tokens, etc.) using configurable sampling strategies, displaying live generation stats and the top candidate symbols as it works. The creator built it with Claude's help by describing the sampling algorithms and having the AI implement them.
CNN sources report that the US military almost intercepted a Chinese vessel in the Middle East this spring after an analyst used an AI chatbot to assess intelligence, which falsely claimed the ship carried nuclear weapons materials. Armed personnel and aircraft were already mobilizing before officials realized the report was based on a fabricated chatbot response and stood down.
CNN reports a Pentagon analyst used an AI chatbot to assess a Chinese vessel's cargo, and the tool falsely concluded it carried nuclear weapons components by blending open-source data with classified signals intelligence. The false finding nearly led US forces to seize the ship before officials realized the assessment was fabricated by the AI.
A US Special Operations Command analyst used an AI chatbot to help analyze intelligence about a Chinese vessel, and the tool wrongly concluded it was carrying nuclear weapons program components through the Middle East. The US military began preparing to intercept and board the ship with air support before officials realized the chatbot had misread the cargo data and the report was baseless.
A new CNBC report finds AI has become deeply embedded in hiring, screening resumes and now conducting first-round interviews. A Greenhouse survey of nearly 3,000 job-seekers across five countries found 63% had gone through an AI-led interview, up 13% from six months prior. These interviews range from voice and text chatbots to digital avatars that mimic human facial expressions.
According to The Information, Meta is building a new smart glasses model called Luna that drops the built-in camera found in its current lineup. Instead, the device relies on six microphones and a side button to let wearers talk directly to Meta's AI chatbot and its Muse AI agent. The glasses could debut as soon as next week's Meta Connect event in Menlo Park.
Meta is reportedly developing a new line of smart glasses, dubbed Luna, that will not include cameras, according to The Information. The devices, expected to launch in October in Clubmaster and Burbank styles, will feature six microphones and ear-facing speakers so users can talk to Meta's AI assistant via a side button, but will lack the recording capability found in Meta's Ray-Ban glasses.
Meta is reportedly building a camera-less version of its Ray-Ban smart glasses, codenamed Luna, following widespread criticism that the camera-equipped models enabled covert recording and harassment. The new glasses would retain six microphones for voice interaction with Meta's AI assistants but drop the camera entirely, and could debut as soon as next week's Meta Connect conference.
According to The Information, Meta is developing a new smart glasses line called Luna that omits the camera found in its current models. The glasses will reportedly come in two styles, Clubmaster and Burbank, and feature six microphones plus ear-facing speakers for interacting with Meta's AI assistant via a side button. Meta may reveal Luna at its Connect conference on September 23-24, with shipping possibly starting in October.
Walt Disney World is rolling out a beta AI assistant on its booking site and app, built on Google's Gemini but limited to official Disney data. Guests will be able to ask conversational questions about hotels, resorts, and amenities, and receive AI-generated summaries on pricing and availability before being directed to booking links. The beta will initially reach randomly selected guests in the US and Canada, excluding Lightning Lane and dining reservations for now.
A viral clip shows ChatGPT's Voice feature confidently miscounting the letter 'e' in 'seventeen,' claiming there are three when there are actually four. When the user, a content creator called Husk, corrected it, the chatbot repeatedly refused to concede the point even after spelling out the word itself. An OpenAI employee later clarified the voice assistant was running on an older model, GPT-Live-1, not GPT-6 as the bot itself claimed, and that it failed to delegate the query to a more capable model.