The AI week at a glance
The key verified events from our daily editions, preserving their original event dates.
AMD Unveils Open Server Blueprints for AI Agents
On October 12, 2026, AMD introduced AMD Helion, an open design blueprint for server racks. The system targets agentic AI, meaning automated software assistants that plan and perform complex multi-step tasks. AMD will share the blueprint with the Open Compute Project, a group that develops shared hardware standards. Data centers can assemble these racks in different ways to fit their needs. One setup uses only central processors, offering up to 32,768 CPU cores to handle agent logic. Another all-in-one option combines standard processors with graphics chips to run AI models on the same machine frame.
Reported facts show AMD wants hardware tailored to multi-step agents rather than simple chat prompts. In our view, open server designs could help companies lower equipment costs and avoid getting locked into one supplier.
UK Government Expands AI Cyber Defences
On October 12, 2026, the British government announced steps to use advanced AI to protect public networks and critical services. The new initiative expands on two earlier test projects. Those trials successfully identified and fixed more than 1,000 security vulnerabilities across government computer systems. Officials also launched preparations for serious incidents involving autonomous AI agents. These are software programs that perform complex tasks with little human supervision. Because agent failures can spread quickly, authorities are examining incident reporting requirements under the upcoming Cyber Security and Resilience Bill.
Governments face fast-moving cyber threats and must adapt their defensive playbooks. In our view, holding automated software accountable before accidents occur is crucial for public trust.