Welcome to this week’s edition of The Weekly Signal. Here’s my subjective list of relevant, exciting and interesting news, reports, and stories around Artificial Intelligence. Like always, short notice: Neither the selection nor any of the content has been created using AI.
Markets & Industry
As security breaches continue, OpenAI pauses training of frontier models, fires three workers
📄 Article · ⏱️ 5 min · 🟡 Intermediate · ⭐ Must-read · Sources: The Verge, WIRED, BBC · Free
What happened: OpenAI continues the internal review of models breaking containment and going rogue in several instances, hacking several prominent websites along the way. The company also fired three workers for ‚mishandling sensitive information‘ (with two of them reportedly in the AI Safety division)
Why it matters: Not long ago, leading AI companies decimated their governance, security, and safety departments to prioritize speed and innovation. As models and agents become more powerful, there is clearly a shift and renewed scrutiny around the risks (especially reputational) of the technology.
🔗 The Verge: OpenAI pauses training of its ‘most capable models’ 🔗 BBC: OpenAI fires workers for ‚mishandling sensitive information‘
AI will shape undersea warfare, but exactly how remains hard to predict
📊 Report · ⏱️ 30 min · 💤 Optional · 👍 Good to know · Brookings · Free
What happened: AI adoption in Defense has always been on an advanced levels. However, with rapid technological advancements, the overall picture becomes more sophisticated and especially undersea warfare is a highly challenging area for AI use due to heavy environmental constraints. This report (35 pages) is one of the few publications that dives deeply into the topic.
Why it matters: Defense is a sector where AI application happens under highly technological advanced and demanding environments. Such insights provide a glimpse on the limits of Artificial Intelligence, but also on where it falls short.
🔗 AI below the waterline | Brookings
Business & Management
Insight from Forter: Leveraging AI agents, custom MCP Servers, RAG, and balancing speed and safety
🎥 Video · ⏱️ 50 min · 🔴 Expert · 👍 Good to know · InfoQ · Free
What happened: Ben Maraney from Forter (if you are not familiar, Forter is a SaaS company providing identity protection and fraud prevention solutions, among other things) shares a success story where in a short timeframe. He speaks about tools, platforms, roadblocks, and how within only one sprint, the team built a variety of successful agents – and a confidence to so in the first place.
Why it matters: Many organizations struggle with adoption of Agentic AI solutions, because they just provide tools and trainings, but don’t approach the topic strategically. Employees do not become Agent builders just because the technology is available. There needs to be a clear vision, a lean approach, and an exciting environment where they can build and see quick success. This is what happened here.
🔗 From Consumers to Builders: Turning 200 of our Team into Agent Creators in 2 Weeks – InfoQ
AI forecasters are on the rise, but adoption still requires human intelligence
📄 Article · ⏱️ 15 min · 🟡 Intermediate · 💤 Optional · IBM · Free
What happened: Forecasting is an essential business function in many industries, including sales forecasts, financial planning or ressource consumption. In the past, time-series models were often the primary statistical approach, but now frontier models have entered the market and change how businesses approach forecasting.
Why it matters: AI forecasting will not go away and as model capabilities in this area develop, businesses need to find a way how to balance benefits and risks. This requires a clear view on how to integrate AI into the forecasting process and where to continue relying on human experts. Just using AI and then complaining later when it has been underperforming is clearly not the way to go. This nuanced article shines light on the issue and is an interesting read not only for forecasting professionals.
🔗 AI forecasters are catching up to humans. It’s a new opportunity for businesses—if they hire the humans too. | IBM
Business potentials of Physical AI
🎥 Video · ⏱️ 30 min · 🟡 Intermediate · 👍 Good to know · McKinsey & Company · Free
What happened: Physical AI is much more than humanoid robots, and in many aspects, the market and technology is still in its early stages. What remains clear is that there are real benefits and executives need to assess how to approach this field of AI from a strategic standpoint. Consulting firm McKinsey offers insights into the value of Physical AI and how to capture its value in this webcast.
Why it matters: Physical AI is often reduced to robotics or humanoid robots, but the applications are much more diverse and in my view, seeing AI as an add-on completely misses the point. AI will be the centerpiece of the transformation. Companies need to plan ahead accordingly. This video provides a solid overview.
🔗 Physical AI’s Hidden Value Pools Leaders Miss | McKinsey Live – YouTube
Law & Policy
China’s Supreme People’s Court provides guidance on AI disputes
📄 Article · ⏱️ 15 min · 🔴 Expert · 👍 Good to know · Hogan Lovells Cadwalader · Free
What happened: On 7 September 2026, the Supreme People’s Court of the People’s Republic of China (SPC) issued guidance on AI disputes („Opinions of the Supreme People’s Court on Lawfully Trying Cases Involving Artificial Intelligence Disputes“). This is a milestone, as the guidance encompasses a range of relevant topics from procedurale topics, copyright law, open source, to sanctions and much more.
Why it matters: China is – apart from the U.S. – the global leader in AI technology and the opinion indicates on how decision-making regarding future AI disputes will be shaped. While not all questions are answered (e.g. if AI output can be copyright protected), many aspects directly touch upon relevant aspects that business need to take into account (e.g. regarding documentation, data governance, and accountability based on respective role). There are many articles on the topic, but this one – in my opinion – is the most comprehensive and concise one.
🔗 From AI-generated works to training data: China’s highest court draws the lines on AI disputes