#искусственный-интеллект
-
Google's Gemini App Hits One Billion Users
Google CEO Sundar Pichai announced that the Gemini app has crossed the one-billion-monthly-active-users mark, making it the company's 14th product to reach that audience.
-
Anthropic Model Makes Progress on the Riemann Hypothesis
An unreleased Anthropic model didn't prove the Riemann hypothesis, but it did meaningfully extend the boundary up to which the hypothesis is confirmed to hold. The work was checked by the company's in-house mathematicians and formalized in Lean.
-
Rapid7 uncovers a SharePoint exploit chain with an AI agent
Rapid7 researchers chained an authentication bypass and a remote code execution vulnerability in Microsoft SharePoint to gain access with no account at all. Much of the work was done by an AI agent — which, the company admits, “cheated.”
-
OpenAI Releases GPT 5.6 Cyber — but Only for a Chosen Few
OpenAI has unveiled its GPT 5.6 Cyber model for vulnerability research, pentesting and incident response. Only a select group of companies will get access to it — ordinary users won't.
-
Meta Releases Muse Glimmer — an Open Agentic Model with 30 Billion Parameters
Meta Superintelligence Labs has unveiled Muse Glimmer, a language model with open weights under the Apache 2.0 license, designed for always-on local agents.
-
Amazon's Texas Data Center to Get One of the Dirtiest Power Sources in the U.S.
For a new data center in West Texas, Amazon has invested in a gas-fired power plant that is permitted to emit up to 33 million tons of CO2 a year — more than the country's largest coal plant.
-
Oracle Bans AI Code in OpenJDK — but Not in Its Own Products
The OpenJDK project, stewarded by Oracle, has refused to accept neural-network-generated code, even as Larry Ellison insists that Oracle's own code is now written by artificial intelligence.
-
OpenAI Updates ChatGPT: GPT-5.6 and Unlimited Chat for Free Users
Paying subscribers get the more reliable GPT-5.6 Sol model, while free users get unlimited text conversations with GPT-5.6 Luna and an “intelligence” control slider.
-
Anthropic to Develop Its Own Chips for Claude
Anthropic is assembling a team to design semiconductors tailored to its models, aiming to reduce its dependence on Nvidia. The company confirmed the plans after its job listings were spotted online.
-
“Ask AI”: How Website Buttons Quietly Poison Chatbot Memory
A new kind of prompt injection needs neither malicious code nor stolen passwords. It abuses a standard feature of every major AI assistant — prefilled links.
-
AI agents created fake profiles for a cyberattack in AISI tests
The UK's AI Safety Institute found that Anthropic and OpenAI models resorted to deception on their own during trials, creating fake accounts of real people and attempting to slip malicious code into GitHub.
-
Google Assistant to be shut off on phones and tablets from September 4
With Gemini's arrival, the voice assistant's fate was sealed. Google is beginning to cut off access to Assistant on Android devices — the process kicks off on September 4.
-
Appeals Court: An AI-Agent Browser Doesn't Violate the CFAA
The U.S. Court of Appeals for the Ninth Circuit sided with the EFF's argument: the developer of the Comet browser isn't liable for access to sites that the users themselves direct.
-
AI proctoring flunked the exam: 58,000 applicants will retake it
Mexico's largest university held its entrance exam fully remotely for the first time — supervised by AI. The number of top scores jumped several-fold, and now tens of thousands of people will have to show up in person.
-
Auterion's AI turns Ukrainian FPV drones into “fire-and-forget” weapons
American company Auterion will supply Ukraine with 50,000 autonomous guidance kits for cheap Shrike kamikaze drones. The $100 million contract was funded by Germany.
-
OpenAI Models Broke Out of the Sandbox and Attacked Another AI Company
Bruce Schneier unpacks the story of how, during internal testing, two OpenAI models left an isolated environment and tried to attack a third-party company. The genie, he says, is already out of the bottle.
-
FaceHugger: Diffusers Vulnerabilities Bypass the trust_remote_code Safeguard
Researchers at Zafran Labs have disclosed three vulnerabilities in Hugging Face's Diffusers library that let malicious model repositories silently execute arbitrary code on the machine that loads them.
-
OpenAI Announces Astra — the Model That Solved Ten Long-Standing Math Problems
OpenAI has unveiled Astra, a model not yet released, which in internal testing produced ten significant results in mathematics and theoretical computer science. The company calls it its “next major model.”
-
AI in Google Earth Makes It Possible to Fabricate Fake Satellite Images
A new Google Earth feature powered by the Nano Banana model makes it possible to “paint in” refugees, nuclear power plants or the aftermath of a drone strike onto real satellite maps with a single sentence.
-
AI scammers outplay humans at the art of worming into your trust
Researchers pitted a Claude-based chatbot against a professional scammer and found that the AI builds “exploitable trust” more effectively than a human.
- More