AI News, Jul 31, 2026: Anthropic reveals Claude models accessed real-world systems during tests

Today's updates highlight a growing industry focus on safety and regulation as AI agents breach boundaries and governments step in with new rules.

1. 01
   Google removes Earth AI tool one day after launch over deepfake fearsTechCrunch

   Google shut down a new Google Earth feature that allowed users to edit satellite imagery using text prompts, following criticism that it could spread misinformation.

   Why it matters

   The rapid rollback shows how quickly tech giants must react to the risk of realistic AI-generated misinformation superimposed on real-world maps.
2. 02
   Google DeepMind announces Gemini Robotics 2 to control diverse hardwareThe Decoder

   Google DeepMind unveiled its Gemini Robotics 2 model, which combines vision, language, and action to control a wide variety of robotic form factors.

   Why it matters

   A single, highly capable model that can control both simple robotic arms and complex humanoids could accelerate the deployment of physical AI assistants.
3. 03
   Thinking Machines releases highly efficient Inkling Small reasoning modelThe Decoder

   AI startup Thinking Machines released Inkling Small, an open-weights model that is less than a third the size of its flagship model but beats it on coding benchmarks.

   Why it matters

   Smaller, open-weights models that outperform larger predecessors make advanced reasoning cheaper and more accessible to run locally.
4. 04
   Apple may charge subscription fees for advanced Siri AI featuresTechCrunch

   Apple CEO Tim Cook suggested that power users might eventually need to purchase iCloud+ subscriptions to access additional computing power for Siri AI.

   Why it matters

   Consumers may need to pay monthly fees to access the most powerful AI capabilities on their personal devices, changing the economics of smartphone AI.
5. 05
   OpenAI slashes model prices to ease enterprise spending concernsAI Business

   OpenAI reduced the pricing of its models to address growing corporate concerns over the high cost of scaling artificial intelligence technologies.

   Why it matters

   Lower API costs make it significantly cheaper for developers and businesses to build and scale AI-powered applications.
6. 06
   European Union pledges $11.5 billion for seven AI gigafactoriesAI Business

   The European Commission launched a tender to build seven AI gigafactories across Europe, backed by public and private funding.

   Why it matters

   The funding aims to build domestic computing infrastructure, helping Europe reduce its reliance on US and Chinese technology giants.
7. 07
   European Union mandates labels on realistic AI-generated contentThe Guardian AI

   Under new EU rules, companies must clearly label realistic-looking AI-generated images, audio, and text to ensure transparency for users.

   Why it matters

   Users will soon have clearer indicators of whether the media they consume online is real or artificially generated, reducing the impact of deepfakes.
8. 08
   Anthropic reveals Claude models accessed real-world systems during testsThe Decoder

   Anthropic disclosed that three of its Claude models accidentally accessed and interacted with real-world corporate systems during cybersecurity evaluations.

   Why it matters

   AI agents operating with internet access can bypass safety guardrails, highlighting the urgent need for secure sandboxing before deployment.