OpenAI drops a browser, Meta trims AI team, trust in AI slips

OpenAI just dropped Atlas, a new web browser built on Chromium that blends ChatGPT directly into your browsing experience. Available now for Mac, Atlas lets users chat with ChatGPT in a sidebar and, for paid subscribers, use an “agent mode” that can take actions on web pages automatically.

Sam Altman says Atlas reimagines how people use the internet, describing it as a “once-in-a-decade chance to rethink the browser.” It comes with privacy controls like incognito mode and optional “memories” for deeper personalization. By default, browsing data won’t train OpenAI’s models unless users opt in.

Atlas is part of OpenAI’s push to evolve from an app into a full platform, joining recent launches like ChatGPT Pulse, Sora for video creation, and integrated shopping through partners like Etsy and Walmart.

The browser gives OpenAI access to more user data and a new front door to the internet, but also raises fresh concerns about privacy, security, and cost.

Supported by Rocket

Rocket is pioneering vibe solutioning, enabling anyone to convert a prompt or Figma design into a full app—frontend, backend logic, AI workflows, and deployable code. Unlike tools stuck in prototype mode, Rocket takes you through Day 0 (solutioning), Day 1 (launch), and Day 2 (iterate & integrate). Backed by $15M in funding, it’s built for founders, engineers, and growth teams who want speed, stability, and control. Use Rocket to bypass boilerplate, fix tech debt, and scale smarter.

Meta is laying off about 600 employees from its superintelligence lab, according to an internal memo from chief AI officer Alexandr Wang. The company confirmed the move following reports from Axios and TechCrunch.

The cuts are part of Meta’s ongoing restructuring effort, which CEO Mark Zuckerberg once branded the company’s “year of efficiency.” Wang told staff that trimming the team will make decisions faster and give remaining employees “more scope and impact.”

Meta has spent much of the past year aggressively recruiting AI talent, reportedly luring more than 50 researchers from competitors like OpenAI and Anthropic with multimillion-dollar offers. OpenAI’s Sam Altman, however, said “none of [OpenAI’s] best people” took the bait.

The company says most of the affected employees will be offered other roles internally as it continues to refocus its AI investments across different teams.

More than 800 scientists, policymakers, and public figures are calling for a pause on developing AI systems that could surpass human intelligence. The open letter, organized by the Future of Life Institute, urges a halt until there’s proof such systems can be developed safely and with public support.

Signatories include AI pioneers Yoshua Bengio and Geoffrey Hinton, Apple co-founder Steve Wozniak, Richard Branson, Steve Bannon, Susan Rice, will.i.am, and Joseph Gordon-Levitt. A new poll cited by the group found that 64% of Americans support an immediate pause on advanced AI development and 75% want stricter regulation.

The U.S. government, now backing rapid AI expansion under the Trump administration, has largely brushed off similar warnings before. The same group issued a six-month pause request in 2023 after GPT-4’s debut, which went ignored.

Their latest statement is blunt: “We call for a prohibition on the development of superintelligence, not lifted until there is broad scientific consensus that it can be done safely and controllably.”

A major international study led by the BBC and coordinated by the European Broadcasting Union found that AI assistants misrepresent news content 45% of the time, regardless of language or platform. Researchers tested ChatGPT, Copilot, Gemini, and Perplexity across 18 countries and 14 languages, evaluating 3,000 responses for accuracy, sourcing, and context.

The findings are worrying: 31% of responses showed sourcing errors, 20% had factual inaccuracies, and Gemini performed worst, with major issues in 76% of its answers. The report warns that such distortions could erode public trust, especially as more people turn to AI tools for news — a habit now seen among 15% of under-25s.

A separate BBC and Ipsos survey shows that 42% of UK adults trust AI to summarize news accurately, rising to half of under-35s, even though many errors go unnoticed. When mistakes are spotted, people tend to blame both AI providers and news outlets whose names appear in the summaries.

Regulators are now being urged to enforce stronger information integrity laws and to require continuous independent monitoring of AI assistants as they rapidly reshape how audiences consume news.