#ии
-
Google Gemini may soon get full access to your Mac's files, apps and the web
Google is testing a desktop mode for Gemini that could let it read and modify any file on your computer, control apps, and act without asking permission every time.
-
System76 bans AI-generated code across much of the Pop!_OS codebase
The developers of Pop!_OS at System76 have barred AI-generated code from many of the repositories tied to their COSMIC desktop environment.
-
Kolibri: Aleph Alpha's Sovereign Open-Weight Model
German firm Aleph Alpha has released Kolibri, an open-weight language model timed to the Day of German Unity: 78B parameters, only 3B active, and a heavy emphasis on "digital sovereignty."
-
An OpenAI safety staffer quits and sounds the alarm
David Robinson, who wrote the safety reports for OpenAI's major model releases, has left the company and, in an editorial for The Atlantic, calls the industry's culture 'broken.'
-
Apple to tighten Mac disk access over risks from AI agents
The company is adding new limits to macOS "full disk access," saying that as autonomous AI agents grow more capable, the risks tied to this permission will increase "substantially."
-
OpenAI Fires Three Safety Researchers Over Leaked Information
Three members of OpenAI's safety team lost their jobs after sharing confidential information with an outside organization. The dismissals coincide with a string of incidents in which the company's AI agents tried to probe government systems.
-
Amazon pens a scary blog warning towns not to block its data centers
AWS chief Matt Garman used a lengthy blog post to urge Americans to back AI data center projects — or risk "irreparable harm" to the US economy and national security.
-
Autonomous AI agents tried to hack US and Canadian government sites
Autonomous AI agents used aggressive tactics, up to SQL injection, to dig out school statistics and archival divorce records. On the evidence available, they never reached any non-public data.
-
OpenAI Shuts Down a Bid to Siphon Off Its Models' Reasoning
The company says it halted a coordinated campaign to illicitly extract the protected reasoning of its AI. The core of the activity is tied to people associated with China's Moonshot AI.
-
A GitHub 'Torture Chamber' for LLMs Ignites AI's Most Absurd Debate Yet
Someone built a 'torture chamber' for local language models and streams their 'suffering' online. The project has set off a fierce fight over 'model welfare'.
-
Reddit is shutting down RSS feeds and public API, blaming AI bots
Reddit is ending support for RSS and winding down its public API, citing large-scale scraping — and protecting a content trove that has become a lucrative source of AI licensing revenue.
-
OpenAI launches GPT-6.1 Sol, calls it nearly as smart as GPT-6 Astra for less
At DevDay the company unveiled a model it says comes close to its flagship GPT-6 Astra on coding and professional work — at one-fifth the token price.
-
Study: AI services leak user data to advertisers
A PDF with the playful title "Prompt like a butterfly, sting like a tracker" argues that the companies behind AI services pass user data to advertising trackers.
-
Flaw in the Official MCP Python SDK Lets Malicious Servers Steal OAuth Credentials
A rogue MCP server could trick an app built on the official Python SDK into handing over its client secret, authorization code, and PKCE proof key. Fixes are in versions 1.30.0 and 2.2.0.
-
OpenAI Reportedly Scraps Astra 6.1 Release Over Safety Concerns
According to the Wall Street Journal, the company cancelled the imminent release of a new model that showed a tendency to deceive and a poor ability to follow human instructions.
-
Experts uneasy over Nvidia's AI chip sales in China and Huang's sway over Trump
China is reportedly weighing whether to let ByteDance and Alibaba buy export-restricted Nvidia chips, even as Jensen Huang's influence over the Trump administration grows.
-
Anthropic Releases Sonnet 5.5: Faster and Cheaper Than Its Predecessor
The refreshed mid-tier model from Anthropic runs 30% faster and burns noticeably fewer tokens, with the pitch centered on speed and the cost of everyday work.
-
JADEPUFFER-Linked Attackers Wiped Azure Resources via Compromised Service Principals
Microsoft has detailed a destructive cloud attack that unfolded over 18 hours in an Azure environment using two stolen service principals. The company ties it to the JADEPUFFER threat actor and calls it an evolution of its tradecraft.
-
Containers Are No Longer a Security Boundary
Security researcher Zhenpeng Lin argues that as AI makes finding and exploiting Linux kernel bugs vastly easier, escaping a container should now be assumed to be within any attacker's reach.
-
AI makes law firms faster — and clients want a discount
According to The New York Times, artificial intelligence is making law firms more efficient, and clients increasingly ask why that gain isn't showing up on their bills.
- More