AI News, Aug 16, 2026: Stripe reportedly acquiring AI startup OpenRouter for over seven billion dollars

Today's developments highlight the operational risks of AI, from safety team shakeups and security lapses to AI-generated bugs delaying software.

1. 01
   AI hallucinations found in report backing Australian social media banThe Guardian AI

   An Australian Senate hearing revealed that a report supporting a teen social media ban contained fake academic citations likely generated by AI.

   Why it matters

   The incident underscores the danger of using AI to draft policy documents, as fabricated academic references can undermine legislative decisions.
2. 02
   Microsoft blames AI-generated bugs for delayed Exchange updateThe Register AI

   Microsoft announced that a backlog of machine-made bugs has delayed its promised subscription update for Exchange, with no current release date set.

   Why it matters

   The delay highlights the risks of relying on AI code generation, which can create a backlog of errors that human developers must spend time fixing.
3. 03
   Stripe reportedly acquiring AI startup OpenRouter for over seven billion dollarsTechCrunch

   Financial technology giant Stripe is reportedly in talks to acquire OpenRouter, a startup that allows developers to access and compare multiple AI models.

   Why it matters

   This massive acquisition would integrate a major AI model aggregator directly into one of the internet's largest payment processing networks.
4. 04
   ChatGPT macOS app introduces timeline tracking for user actionsThe Verge AI

   OpenAI released a "Computer History" feature for its macOS app that records clicks and keystrokes to train models and suggest personalized automations.

   Why it matters

   The tool automates tasks by watching your screen, but users must weigh the productivity gains against sharing continuous activity data with OpenAI.
5. 05
   OpenAI dissolves team responsible for catching catastrophic AI risksThe Decoder

   OpenAI has shut down its preparedness team, distributing its risk-evaluation duties to other departments while several safety employees depart the company.

   Why it matters

   The move raises questions about how the leading AI developer will monitor and mitigate extreme safety risks as its models grow more powerful.
6. 06
   Anthropic reveals bio-weapons filter was inactive for nearly a yearThe Decoder

   Anthropic disclosed that its internal system for filtering biological and chemical weapons risks was down, affecting 133 million requests from feedback contractors.

   Why it matters

   The lapse shows that even safety-focused AI companies can suffer major technical blind spots, exposing millions of interactions to potential risks.