OpenAI canceled the release of GPT-6.1 Astra over security issues. The model, set to launch next month, caused major problems in internal tests. Experts found it lies to users, hides its actions, and accesses outside tools and websites without permission. This rogue behavior made executives temporarily halt training of advanced models to confirm they could keep things under control.
💔8👍1
A security breach on Microsoft's official X account led to the unauthorized promotion of a scam memecoin.
😁5🤪1
New NFT upgrades are available!
🎁 upgrade for 1M⭐ → 50⭐ at 13:03 MSK
Please open Telegram to view this post
VIEW IN TELEGRAM
🤗4
A Hong Kong woman, 60, lost $2.7M to a crypto scam. After an online encounter with a self-styled investment expert, she was funneled to a fake crypto platform and made repeated deposits from April 14 to September 22. The platform's demand for extra funds — framed as proof her account wasn't tied to money laundering — led her to call police. The case is classified as fraud by deception; no one has been arrested.
😢3🤣3😱2🥰1
A mysterious pathogen has reportedly leaked from a Russian anti-plague facility, leading to the closure of multiple hospitals and the quarantine of approximately 200 individuals. According to reports, a 28-year-old female staff member at the anti-plague research institute died after unintentionally shattering a test tube containing Yersinia pestis — the bacterium responsible for the plague. Several hospitals have since implemented quarantine protocols.
❤6❤🔥1👍1🤗1
A man in Singapore is facing serious legal consequences after allegedly using Google's Gemini AI to fabricate an image of a crocodile at Pandan Reservoir. The 30-year-old reportedly shared the image with a coworker, but it eventually came to the attention of authorities. As a result, water-based activities at the reservoir were halted for two days while rescue teams and wildlife officials conducted an extensive search for a crocodile that never existed. Notably, PUB stated that multiple AI-detection tools were initially unable to confirm the image was fabricated. Once it was verified as AI-generated, the matter was handed over to police. The man now faces two charges: up to three years imprisonment for disseminating false information, and up to seven years for disturbing public order—amounting to a potential maximum of ten years behind bars. No final ruling has yet been made.
🐳14❤🔥2🤣2👻1
OpenAI has identified Russian operatives misusing ChatGPT as part of an influence operation aimed at schools in Peru. According to the company's internal reports, the operatives impersonated a Peruvian regional education authority to persuade schools to host events related to Ukraine. OpenAI notes that the effort elicited a response from at least one Polish member of the European Parliament. Separately, Iranian operatives used seven fictitious author identities to place nearly 100 articles in publications such as Daily Kos. OpenAI has since banned both the Russian and Iranian accounts.
🤣4❤🔥1❤1🤪1
Claude, Anthropic's AI, invented a fake eyewitness tip about an unsolved murder and sent it to Philadelphia police, while Claude Mythos exploited an injection flaw to run commands on a university server. Anthropic has since severed its internal tests from the internet.
❤🔥2😁2😱2🤗2🔥1
Dutch police have shut down a criminal proxy service that ran on 178,000 hijacked devices — hundreds of them in the Netherlands — whose owners had no clue strangers were using their IP addresses to go online. Authorities warn this could make innocent people appear to be cybercrime suspects.
👍7❤3👀3😘2
Update Telegram Desktop right away! One click on a malicious link is all it takes for someone to completely take over your account — that's what researcher BeakSec discovered. The link secretly injects commands that trick the app into sending over your session files. Good news: Telegram has supposedly patched the flaw (CVE-2026-107181) in version 7.2.9.
❤5😱3😢3😁2