Nvidia's security gap exposed, Italy fines Meta, EU sets an AI Act compliance shortcut
Reporting shows Nvidia's security leadership changed hands months before the Huang scandal broke. Italy fines Meta over 800 million euros, and the EU publishes its first official shortcut for AI Act compliance.
Every line links to a primary source in the full briefing.
01Nvidia's security leadership turned over months before the Huang scandal
Nvidia's long-serving chief security officer left the company in August, months before CEO Jensen Huang stepped down over personal relationships that drew scrutiny, including one tied to a company under investigation for allegedly funneling Nvidia chips to China. Reporting raises the question of whether anyone inside Nvidia's security function flagged concerns earlier, though nothing has been confirmed linking the two departures.
Leadership scandals rarely stay contained to the executive suite: they tend to surface gaps in internal oversight that can affect suppliers and partners downstream.
Source: TechCrunch
02US lab says Chinese models trained on sanctioned Nvidia chips
A US research lab reports that Chinese AI models appear to have been trained using Nvidia's Blackwell chips, despite export restrictions meant to block that access. The claim adds to ongoing questions about how effectively chip sanctions are being enforced.
Weak enforcement of chip export controls could keep pricing and supply uncertain for any business relying on AI hardware availability.
Source: Chosunbiz
03Italy fines Meta over 800 million euros for its photo gallery feature
Italian regulators fined Meta roughly 805 million euros over practices tied to a photo gallery feature. The penalty is one of the larger regulatory actions against the company in Europe this year.
It is a reminder that European regulators are actively enforcing against big platforms, not just drafting rules, which raises the bar for any company handling user data or AI-driven features.
Source: El Output
04EU publishes first official shortcut for AI Act compliance
The European Commission published the first harmonised standard, EN 18286, supporting the AI Act in the Official Journal. Companies that build a quality management system around this standard get a legal presumption of conformity with Article 17 requirements for high-risk AI systems, and it is designed to layer on top of existing standards like ISO 9001 rather than replace them.
Businesses deploying high-risk AI systems in the EU now have a documented, recognised path to prove compliance instead of guessing what regulators will accept.
Source: European Commission
05Runway open-sources a video model that controls robots
Runway released Praxis-1, a model that uses knowledge from video training to predict robot movements from camera images and text instructions, under an open Apache 2.0 license that allows commercial use. It leads an industry robotics benchmark but still cannot plan complex multi-step actions on its own.
Cheaper, open robotics software lowers the barrier for smaller manufacturers or logistics operators to experiment with automation without building custom datasets.
Source: The Decoder
Check whether your AI vendor contracts reference recognised compliance standards
With the EU's first official AI Act compliance standard now published, ask any AI vendor serving your business whether they plan to align with it. Getting ahead of this now avoids scrambling later when enforcement ramps up.
Sources
5 links- 01Multa a Meta en Italia: 805 millones por la galeríaEl Output
- 02First harmonised standard supporting the AI Act published in the Official JournalEuropean Commission
- 03Where was Nvidia’s security team?TechCrunch
- 04U.S. lab says Chinese AI models trained on Nvidia Blackwell chips despite sanctionsChosunbiz
- 05Runway's open-source Praxis-1 turns video AI into a robot controllerThe Decoder