AI News
AI news that matters. Updated daily.
No stories match your filters.
Anthropic Withheld Claude Mythos for Safety - Then Unauthorized Users Got It Anyway
Three weeks ago, Anthropic was telling the world that Claude Mythos - its most capable model - was so dangerous it couldn't be released publicly. The cybersecurity capabilities were too potent, the risks too real. Now, according to Bloomberg, a small group of unauthorized users has had access to that same model anyway.
Claude Now Connects to Spotify, Uber Eats, TurboTax, and Other Personal Apps
Anthropic just expanded Claude's app connectors beyond workplace software into the apps people actually use off-the-clock. The new lineup includes Spotify, Uber Eats, Audible, Instacart, AllTrails, TripAdvisor, TurboTax, and others.
Sierra Acquires YC-Backed French Startup Fragment
Sierra, the AI customer service startup co-founded by Bret Taylor, has acquired Fragment, a Y Combinator-backed French startup. No acquisition price was disclosed.
Claude Desktop Quietly Installs a Browser Communication Bridge Without Clear Disclosure
When you install Claude for Desktop, you get more than an AI chat window. Security researchers have found that the app installs a native messaging bridge - a system-level component that lets Claude communicate directly with your browser - without clearly disclosing this during setup.
Meta Cuts 10% of Workforce, Eliminating Around 8,000 Jobs in May
8,000 jobs. That's how many Meta employees will be cut in May, according to an internal memo from chief people officer Janelle Gale, published by Bloomberg. The cuts represent roughly 10% of Meta's total workforce. Alongside the layoffs, Meta is also canceling approximately 6,000 open positions that were never filled - so the real headcount reduction is closer to 14,000 roles gone from the org chart.
AWS Names Autonomous Long-Running Agents the Next Enterprise AI Shift
Three years of enterprise chatbot rollouts have produced a consistent lesson: answering questions is useful, but finishing tasks is where the real value sits.
Anthropic Publishes Same-Day Postmortem on Claude Code Quality Reports
Most AI companies respond to quality complaints with vague posts about "ongoing improvements." Anthropic took a different approach on April 23: a formal engineering postmortem documenting recent quality issues with Claude Code.
Microsoft to Invest $18B in AI Infrastructure Across Australia
$18 billion. That's Microsoft's planned commitment to AI infrastructure in Australia, making it one of the largest single-country AI infrastructure bets the company has announced.
AI Labs Are Cutting Off Heavy Users as Token Costs Outpace Revenue
Earlier this month, millions of OpenClaw users discovered their app had changed overnight. The viral AI agent tool - one of the fastest-growing software products of 2026 - had been severely restricted by Anthropic, the company whose AI models power it. Features disappeared. Limits dropped. No warning, no grace period.
Microsoft Rolls Out Agent Mode in Word, Excel, and PowerPoint
"Vibe working" was Microsoft's marketing phrase for it. The actual feature, now rolling out, is called Agent Mode - and it's landing in Word, Excel, and PowerPoint this week.
OpenAI Launches GPT-5.5, Targeting Complex Coding and Research Workflows
OpenAI launched GPT-5.5 on April 23, billing it as their most capable model to date - faster than GPT-5 and built specifically for demanding, multi-step work: coding, research, and data analysis that spans multiple tools in a single session.
Ars Technica Publishes Formal Newsroom AI Policy
Note: The source article at arstechnica.com was inaccessible during production. The article content below cannot be written with factual accuracy without the original text. Please provide the article content so this can be rewritten with specific policy details, quotes, and rules.