Daily AI News | NVIDIA and MediaTek expand their partnership across cloud, edge, and automotive AI
# Daily AI News | NVIDIA and MediaTek expand their partnership across cloud, edge, and automotive AI
## Lead Story
### NVIDIA and MediaTek expand their partnership across cloud, edge, and automotive AI
NVIDIA and MediaTek said they are deepening their long-running partnership to build AI computing platforms spanning data centers, local devices, and vehicles. NVIDIA confirmed the platform scope, while TechCrunch reported that NVIDIA will also invest $3.5 billion in MediaTek. Product timelines and execution details remain to be disclosed.
**AI take:** The strategic point is not one chip, but NVIDIA extending a common AI ecosystem across cloud, devices, and cars.
## More News
### OpenAI expands ChatGPT Ads self-service access and cites a $1 billion annualized run rate
OpenAI said ChatGPT Ads has reached a $1 billion annualized revenue run rate and expanded self-service buying to more international markets. This is a new stage beyond the earlier India rollout. The announcement does not promise simultaneous availability everywhere or change stated subscription entitlements.
**AI take:** The ad business is moving from pilot to scale; the next test is balancing monetization with the user experience.
### The Pentagon adds secured ChatGPT and Grok versions to GenAI.mil
The Pentagon has added secured versions of ChatGPT and Grok to GenAI.mil, its central generative-AI portal, alongside Gemini, TechCrunch reported. The deployment is within a controlled government environment and does not mean public versions may handle every category of sensitive information.
**AI take:** The meaningful shift is a governed multi-model gateway for defense work, not merely another chatbot purchase.
### Codex and ChatGPT Work reach 25 million active users and reset paid usage
OpenAI Codex lead Tibo Thsottiaux said Codex and ChatGPT Work reached 25 million active users and reset usage for every paid subscription. The post confirms a fresh allowance, not a permanent increase to plan limits, and gives no universal numeric uplift.
**AI take:** The reset restores near-term capacity, but long-running work should still be planned around the limits shown in each account.
### Instagram starts limiting reach for undisclosed AI-generated profiles
Instagram has begun limiting distribution for AI-generated profiles that do not disclose their nature, TechCrunch reported. The measure connects disclosure requirements to a practical reach penalty. It targets undisclosed AI profiles rather than banning every virtual character or all AI-assisted content.
**AI take:** Platform governance is moving from labels to distribution consequences, making identity disclosure a baseline operating requirement.
### Didi begins driverless passenger testing of its R2 robotaxi in Beijing and Guangzhou
Didi has begun driverless passenger testing of its next-generation R2 robotaxi in designated areas of Beijing and Guangzhou. The vehicle uses Didi’s autonomy software, 33 sensors, and a three-domain compute platform. This remains a bounded test, not citywide commercial service.
**AI take:** Moving into passenger tests matters, but scale, intervention rates, and permits will determine commercial readiness.
### Microsoft Research releases GigaPath-Flash and GigaTIME-Flash for efficient pathology studies
Microsoft Research released GigaPath-Flash and GigaTIME-Flash, pathology foundation-model variants designed to reduce compute while retaining reported performance for imaging and spatiotemporal analysis at larger population scale. They are research releases; clinical use still requires independent validation, regulatory review, and deployment assessment.
**AI take:** For medical AI, reducing compute enough to support real population-scale studies can matter as much as headline accuracy.
### Ruijin Hospital and Huawei Cloud release the RuiPath 2.0 pathology model
Ruijin Hospital and Huawei Cloud released RuiPath 2.0 with upgrades for pathology-assisted diagnosis, explainability, rare-disease training, and lighter deployment. It targets clinical pathology workflows, but a model release does not establish independent validation or regulatory clearance for every use case.
**AI take:** Lightweight deployment and explainability are closer to real hospital adoption barriers than one benchmark score.
### Hisense launches JUOS as an AI operating system for TVs, projectors, and appliances
Hisense launched JUOS, a household AI operating system built around its Xinghai model and a system-level assistant, to connect televisions, projectors, and other appliances. The release aims to move from one-off commands to cross-device coordination, while device coverage, upgrade timing, and long-term compatibility remain to be tested.
**AI take:** Home AI needs more than another voice interface; shared state, permissions, and durable compatibility are the hard parts.
### Meeting assistant Circleback adds a free tier with paid plans starting at $14
Circleback introduced a free tier and revised paid plans starting at $14 per month, TechCrunch reported. The free tier lowers the barrier to trying transcription and meeting notes. Users should still check the current product terms for feature allowances and team requirements.
**AI take:** A free tier can widen acquisition, but accuracy, meeting-platform coverage, and data governance remain the durable differentiators.
## Rumors
