Alibaba researchers introduced Hierarchical Decoupled Policy Optimization (HDPO), a reinforcement learning framework that trains AI agents to balance task accuracy and execution efficiency.Salesforce launched Agentforce Operations, a workflow platform that turns back-office workflows into tasks for specialized AI agents. The platform breaks down uploaded processes or Salesforce Blueprints for agents to execute.xAI shipped a new proprietary base LLM, Grok 4.3, and a new voice cloning suite. Grok 4.3 features permanently enabled reasoning, a 1 million-token context window, and API pricing of $1.25/M input tokens and $2.50/M output tokens.LlamaIndex CEO Jerry Liu says the scaffolding layer that developers once needed to ship LLM applications -- indexing layers, query engines, retrieval pipelines -- is collapsing.The Musk v. Altman court battle over the future of OpenAI will be streamed live via YouTube next week. The stream is expected to run between 11AM and 5PM ET while court is in session. Recording and rebroadcasting the stream is prohibited.At TechCrunch's StrictlyVC event, Replit CEO Amjad Masad discussed the company's trajectory from $2.8M revenue in 2024 toward a billion-dollar run rate.The U.S. Defense Department signed deals with Nvidia, Microsoft, Amazon Web Services, and Reflection AI to deploy their AI tech and models on classified networks for "lawful operational use." This follows earlier agreements with Google, SpaceX, and OpenAI.The Pentagon has struck deals with OpenAI, Google, Microsoft, Amazon, Nvidia, Elon Musk's xAI, and startup Reflection for AI use in classified settings, leaving out Anthropic after declaring it a supply-chain risk.Minnesota became the first state to pass a law banning nudification apps that make it easy to "undress" images of real people. Developers face extensive damages and fines up to $500,000 per flagged fake AI nude. The law exempts products requiring "technical skill" to nudify.Musk testified day 3 in his OpenAI lawsuit, claiming delayed filing after believing OpenAI's Microsoft deal didn't violate its non-profit mission. He called xAI the smallest AI player, trailing Anthropic, OpenAI, Google.OpenAI has restricted access to its Cyber security tool after Elon Musk publicly criticized Anthropic's Mythos security model. The move highlights the ongoing tensions between Musk and OpenAI leadership, as well as the competitive dynamics in the AI security space.Elon Musk spent three days testifying as the first witness in his trial against OpenAI but stumbled at least seven times in ways that possibly put his chances at winning in jeopardy.Meta ended its contract with Sama, a Kenya-based data annotation firm, less than two months after Sama workers reported viewing sensitive footage recorded by Ray-Ban Meta smart glasses.Legal AI startup Legora has reached a $5.6 billion valuation in its latest funding round, intensifying competition with Harvey, the other major player in the legal AI space. The funding round highlights the growing investor confidence in AI-powered legal technology.The Verge's podcast coverage of the Musk v. Altman trial highlights Musk's difficult week in court, where his testimony appeared "petty" and unprepared compared to his prior claims of being the "actual driving force" behind OpenAI.Secretary of Defense Pete Hegseth publicly labeled Anthropic CEO Dario Amodei an "ideological lunatic" during a Senate Armed Services Committee hearing, escalating political tensions over AI safety requirements.The jury in the Musk v. Altman trial was dismissed early to allow OpenAI's legal team to handle an objection to Jared Birchall's testimony about xAI's bid for OpenAI assets.The Musk v. Altman trial has revealed extensive evidence including emails dating back to 2015 that provide insight into OpenAI's foundations and early tensions.During the Musk v. Altman trial, Jared Birchall, Musk's finance chief, made a surprising statement about the xAI bid for OpenAI's assets while the jury was dismissed.