RESOLVING THE LEDGER
This is one outlet's own report from The Register — the article as it was filed. Other outlets are covering the same event; open the full story to compare every source side by side.
See the full story · 1 sourcesConventional chatbots respond to individual prompts and often retain only limited context. AI agents may carry out longer, multi-step jobs that span sessions, making persistent memory more important. Object storage vendor MinIO claims it can provide that memory. AI agents can help make decisions, create documents, draft analyses, and answer questions once handled by people alone. MinIO characterizes the resulting output as organizational knowledge – a record of what an enterprise knows, decides, and learns – and argues that controlling it will be essential to scaling agentic AI. AIStor is MinIO's object storage software. AIStor Memory treats agent memory as a native data type alongside objects and tables, allowing agents to retain context across sessions, resume interrupted work, and operate on enterprise data under existing governance controls. MinIO says the resulting knowledge remains on infrastructure controlled by the customer. AB Periasamy, MinIO co-founder and co-CEO, said: "Knowledge generated by AI agents becomes organizational memory, and organizational memory belongs on enterprise-controlled infrastructure. AIStor Memory brings long-term memory, persistent workspaces, and secrets together on a single enterprise-controlled foundation. A single agent's memory becomes the shared substrate for the entire organization." MinIO says AI teams currently have to combine object storage, vector stores, metadata databases, secrets managers, governance tools, and synchronization pipelines to give agents persistent memory. AIStor Memory promises an integrated alternative that mounts directly into existing sandboxes and works with current tools and frameworks without modification. AIStor Memory is built for long-running, multi-step agent workflows, where work must survive interruption and sensitive enterprise data must stay governed. MinIO cites four example use cases: Software engineering agents working across large codebases. Deep research and analysis spanning hours or days. Human-in-the-loop workflows that pause and resume over extended periods. Enterprise AI systems handling governed or regulated data. AIStor Memory sits beneath the KV-cache layer – served by MinIO's MemKV – and preserves the broader state of an agent's work rather than only the context used during inference. MinIO lists the following capabilities: "Infinite" context. Memory scales with available capacity rather than the model's context window, without data being truncated, summarized, or evicted from the persistent store. Standard interfaces. Memory, Workspace, and Vault are available over HTTPS or a POSIX folder mount. Integrated, not assembled. No separate database, vector store, metadata tier, or synchronization pipeline. Enterprise-grade durability. Memory uses AIStor's erasure coding, bitrot protection, encryption, compression, and fault-tolerance features to protect against drive, rack, and datacenter failures. No memory leakage. Memory stays on customer-owned infrastructure, protected by customer-held keys, rather than leaving the organization's environment. MinIO says that as agent-generated work accumulates, memory created for individual agents can become a shared organizational record built by humans and machines. The question, it argues, is where that record will reside and who will control it. AIStor Memory is the company's answer: keep it on enterprise-controlled infrastructure. Bootnote The notion of organizations having AI activity-related memories is spreading, with data protector and resilience supplier HYCU and IT Brand Pulse separately surfacing the concept. ®
AIPROPX is an independent multi-source news index — we track, compare, and connect coverage from across the web into one place you won't find anywhere else.