AI and cloud briefing: an unpatched GeoServer zero-day, DeepSeek's V4 Pro, and Amazon's $220bn AI bet

Here is our latest round-up of what’s changing in AI and cloud for UK businesses, with links to the original sources so you can read further.

Lead story: unpatched GeoServer zero-day under active attack

A newly disclosed zero-day flaw in GeoServer, the widely used open-source geospatial data platform, is already seeing active exploitation attempts. The vulnerability, an SQL injection flaw in GeoServer’s jsonArrayContains function, was first disclosed on 12 August 2026 by a security researcher on X. Threat intelligence firm watchTowr says it observed exploitation attempts within hours of the disclosure, with hundreds of probes traced back to a small pool of IP addresses. No CVE identifier has been assigned yet, and no patch is available. Under certain configurations, the flaw can lead to full remote code execution, and GeoServer has a track record of being targeted at scale once vulnerabilities become public. Full detail is available from The Hacker News and SecurityWeek.

An unpatched SQL injection zero-day in GeoServer is being actively probed for remote code execution, with hundreds of exploitation attempts recorded within hours of disclosure

Why this matters: attackers moved from disclosure to active scanning in a matter of hours, which leaves very little time to react once a flaw like this becomes public. If your business or any supplier runs GeoServer, it is worth checking with your IT team whether any instances are internet-facing, restricting public access where possible, and watching for a vendor patch.

AI: DeepSeek ships V4 Pro as its flagship model leaves preview

DeepSeek has released the general availability version of its flagship model, DeepSeek-V4-Pro-0813, moving it out of preview on 13 August 2026. It’s a mixture-of-experts model with 1.6 trillion total parameters and 49 billion active per token, supporting a context window of just over 1 million tokens and up to 384,000 tokens of output. The release focuses heavily on agent capabilities: tasks where the model uses tools, executes code, and completes multi-step workflows with limited human input. It’s available now through DeepSeek’s API, app and web channels, though the company has also signalled that API prices are set to rise, and by a significant margin. Coverage is available from Unite.AI and Artificial Analysis.

DeepSeek ships V4 Pro, a 1.6 trillion parameter model with a 1 million token context window built for agentic tasks

Why this matters: the continued focus on agent capabilities, rather than just raw benchmark scores, reflects where most model providers are now competing. It’s a useful model to watch if your business is evaluating tools that use AI to complete multi-step tasks, though the flagged price increase is worth factoring into any cost comparison before committing.

Cloud: Amazon raises 2026 AI infrastructure spending to $220 billion

On its second-quarter earnings call, Amazon raised its 2026 capital expenditure outlook to around $220 billion, up from roughly $200 billion, citing higher memory-chip costs. Despite the increase, CEO Andy Jassy said Amazon still expects AI capacity to remain constrained through 2027, with contracted demand already stretching into 2028. AWS itself posted its fastest revenue growth in 18 quarters, up 36.7% year on year to $42.2 billion, with the business now running at a $169 billion annualised revenue rate. Amazon also disclosed an AWS backlog of $496 billion, growing at a triple-digit rate. Full detail is available from Data Center Knowledge.

Amazon raises its 2026 AI infrastructure spending to around $220 billion, with AWS revenue growing at its fastest pace in 18 quarters

Why this matters: this level of sustained spend, alongside a backlog that already extends to 2028, points to continued tight capacity and ongoing competition for compute among the major cloud providers over the next few years. For UK businesses planning AI workloads, that is a reason to book capacity and pricing early where possible, rather than assuming availability will simply be there when needed.

The takeaway

Today’s stories cover three different pressures businesses are facing at once: a reminder that unpatched infrastructure can be found and probed within hours, not days; a fast-moving frontier of AI models increasingly built around agent capabilities rather than just raw scale; and a cloud market where demand for AI infrastructure keeps outpacing supply. If your business runs GeoServer anywhere in its stack, that’s the item to check first today. If you’d like help reviewing your infrastructure exposure or your AI and cloud roadmap, get in touch.

AI Agent based research and content, AI can do mistake, please help us to improve.