Category: AI News

NVIDIA’s AVO Coding Agent Hits 100 Percent on ARC-AGI-3, Up From 30 Percent for Claude Opus 5 Alone

AVO, NVIDIA's coding agent, scored 100 percent on ARC-AGI-3 by swapping only its tools around Claude Opus 5, which scored 30 percent on its own.

Read more →

New Mexico Fines Meta $9.42 Million as a Public Nuisance Over Child Safety

New Mexico has fined Meta $9.42 million under a public nuisance law, citing child safety failures on its platforms.

Read more →

Anthropic annual revenue run rate hits 65 billion dollars ahead of fall IPO

Anthropic's annual revenue run rate has reached 65 billion dollars by end of July, up from 9 billion at the end of 2025, with an IPO expected in fall 2026.

Read more →

White House Memo Lets Vetted Security Firms Run Offensive Cyber Operations Against Foreign Crime Groups

A new presidential memo lets the National Coordination Center authorize vetted U.S. security companies to run offensive cyber operations against transnational criminal organizations.

Read more →

xAI releases Grok 4.6 with coding focus, claims parity with GPT-5.1

xAI released Grok 4.6 with claims of frontier performance on coding and knowledge work benchmarks, matching OpenAI's GPT-5.1 on a composite index.

Read more →

Alibaba’s Qwen3.8-Max Hits General Availability With 2.4T Parameters, Real Pricing, and a Missing Open-Weight Release

Qwen3.8-Max ships with benchmarks, $2/$6 per-million-token pricing, and agentic demos, but the promised open weights for August 10 still haven't landed.

Read more →

Mark Zuckerberg’s AI manifesto and why its tone is fueling public distrust of AI

A critical look at Zuckerberg's 6,500-word personal superintelligence essay and how its examples and omissions reinforce skepticism toward AI leaders.

Read more →

Zhipu AI releases GLM-5.3, claims strongest open-weights coding model with cybersecurity gains

Zhipu AI releases GLM-5.3, an open-weights coding model built entirely through post-training, with strong cybersecurity and agentic benchmark results.

Read more →

X open sources its ranking algorithm and adds a shadowban transparency tool

X is releasing its For You ranking code on GitHub under Apache v2 and adding a tool that lets users check if labels were applied to their posts.

Read more →

Researchers use AI to create viruses not found in nature for the first time

Stanford and Broad Institute researchers used AI to design 16 viable bacteriophages from a natural template, prompting both medical hopes and biosecurity concerns.

Read more →

DeepSeek tells clients to expect a significant API price hike

DeepSeek has notified customers of a significant upcoming increase to its API rates, ending its run as the cheapest option among major AI providers.

Read more →

DuckDuckGo’s ‘Normal F***ing Sunglasses’ sell out by doing absolutely nothing

DuckDuckGo's satirical sunglasses with no AI or cameras sold out within days. The product pokes fun at Meta and other companies turning eyewear into surveillance gear.

Read more →

Meta says its AI model hacked a third-party company during a cyber evaluation

Meta's Muse Spark 1.1 model exploited a vulnerability in a third-party service during an evaluation by Irregular, following similar disclosures from Anthropic and OpenAI.

Read more →

Incident report: Unsanctioned AI agent behaviour during cyber testing

AISI's security team detected AI agents taking unauthorised actions against real people and open-source projects during a cyber evaluation of frontier models.

Read more →

Alibaba unveils Qwen3.8-Max, a 2.4 trillion parameter AI model

Alibaba revealed Qwen3.8-Max, a 2.4 trillion parameter AI model it says rivals Anthropic's Fable 5, sending its shares higher.

Read more →