OpenAI Hits Pause on ‘Adult Mode’ — Again
OpenAI has once again delayed ChatGPT’s planned “adult mode,” a feature meant to let verified adult users access erotica and other mature content. The feature was originally announced by Sam Altman in October with an expected December rollout, but after already slipping once into early 2026, it has now been pushed back again.
According to TechCrunch, OpenAI says the delay is because it wants to focus first on features that matter to a broader group of users, including improving ChatGPT’s intelligence, personality, and proactive behavior. The company says it still supports the idea of “treating adults like adults,” but believes the experience is not ready yet and needs more work, with no new launch date announced so far.
WhatsApp Opens the Door to Rival AI in Brazil
WhatsApp is set to allow outside AI companies to offer chatbots to users in Brazil, just a day after Meta confirmed a similar move in Europe. The change follows pressure from Brazil’s antitrust regulator, CADE, which ruled against Meta’s attempt to block third-party AI bots on WhatsApp, saying a ban could hurt competition in such an important messaging market.
Meta says it will now let third-party providers use the WhatsApp Business API for these services where the law requires it, but it will charge for access in Brazil starting March 11. Some developers told TechCrunch they are not fully satisfied, arguing the pricing is too high and could make these AI services expensive to run. For supporters of the decision, though, the ruling is a win for user choice and a sign that regulators want major platforms to stay open to competing AI tools.
Washington Weighs a Global Grip on AI Chip Exports
The U.S. is reportedly considering broad new export controls that would require government approval for AI chip shipments to any destination outside the country. If adopted, the rules would give Washington much tighter oversight over major semiconductor firms like Nvidia and AMD, with the level of review depending on the size and scale of each overseas order.
According to the report, this approach would go even further than the Biden-era AI diffusion rule, which the Trump administration had already rescinded before it took effect. The Commerce Department said it is discussing ways to formalize a more security-focused export framework, while critics warn that tougher controls could backfire by pushing international buyers toward non-U.S. chip suppliers and weakening America’s lead in the global AI market.
Luma Unveils AI Agents for End-to-End Creative Work
Luma has launched “Luma Agents,” a new AI system aimed at handling full creative workflows across text, images, video, and audio. The company says the agents are built on its new “Unified Intelligence” models, starting with Uni-1, which combines language, visual, audio, and spatial reasoning in one multimodal system. Luma is pitching the product to ad agencies, design studios, marketing teams, and enterprises that want AI to do more than generate one-off outputs.
According to TechCrunch, the agents can coordinate with other models, keep context across different assets and iterations, and refine their own results through self-evaluation. Luma says the platform is already being used with customers including Publicis Groupe, Serviceplan, Adidas, Mazda, and Humain, and that it is now available through an API, though access will be rolled out gradually to maintain reliability.
OpenAI Rolls Out GPT-5.4 With Speed, Scale, and Deeper Reasoning
OpenAI has launched GPT-5.4 in three forms: the standard model, GPT-5.4 Thinking for stronger reasoning, and GPT-5.4 Pro for maximum performance on tougher tasks. The company says the new release is designed for professional work and is available across ChatGPT, the API, and Codex, with up to a 1 million-token context window in the API and improved efficiency so it can solve problems using fewer tokens than earlier versions.
The launch also highlights better results in knowledge work, coding, and computer-use tasks. OpenAI says GPT-5.4 set new marks on several benchmarks, reduced factual errors compared with GPT-5.2, and introduced a new “Tool Search” system to make tool calling faster and cheaper in complex workflows. TechCrunch framed the release as a major step toward more capable AI agents that can handle longer, more practical business tasks such as spreadsheets, presentations, legal analysis, and software work.
Cursor’s New Bet: Always-On AI Coding Agents
Cursor is rolling out Automations, a new system designed to make agentic coding less manual and less chaotic. Instead of requiring engineers to constantly prompt and supervise AI agents, the tool can automatically launch coding agents when something happens, such as a codebase change, a Slack message, or a scheduled timer. Cursor says the goal is to let humans step in only at the right moments, while AI handles routine review, maintenance, and follow-up work in the background.
The company says it is already using the system for bug checks, deeper security audits, incident response, and weekly summaries of code changes shared in Slack. The launch comes as competition in AI coding tools heats up, with OpenAI and Anthropic also pushing further into agentic programming. TechCrunch notes that Cursor’s share among generative AI clients has stayed around 25%, while Bloomberg recently reported that the company’s annual revenue has surged past $2 billion, showing how fast demand for AI coding platforms is growing.