Agentic AI

New watsonx Orchestrate capabilities are designed to discover third-party agents and assign each agent an identity that can be governed through existing enterprise identity systems.

Apple plans additional controls for Full Disk Access in macOS as autonomous agents increase privacy risks. Enterprise teams should review broad permissions, informed consent and revocation before deployment.

The new model is positioned as a faster, lower-cost complement to Opus 5.5, while higher cyber capability brings stronger safeguards and a fresh governance trade-off.

Agentic FNOL is designed to capture, validate, enrich and route claims across digital, voice and mobile channels, bringing governance questions to the first contact with a policyholder.

Mastercard is testing a probability score for AI-initiated payments and combining identity, intent, behaviour and fraud signals to support agentic-commerce authorisation decisions.

RSA Agent ID applies identity-governance controls to autonomous agents, including discovery, lifecycle management and human approval for high-risk actions.

New Nexus tools are designed to pre-investigate cases, assemble entity context and draft narratives while keeping the investigator responsible for the final decision.

OpenShell and the Sentry reference design move agent monitoring into runtime and hardware layers, with JPMorganChase, Microsoft and leading security vendors among the named supporters.

Amazon Web Services argues that prompt injection, excessive agency and data leakage share one root problem: organisations have not controlled what an agent may trust, retrieve or do.

BNP Paribas plans Gemini-powered agents for credit memos and other CIB workflows, with authenticated access, monitored interactions and limits on which data can enter the public cloud.

Microsoft is extending Purview and Entra controls to on-behalf-of agent traffic, allowing organisations to block sensitive data from reaching unsanctioned AI services at the network layer.

Darktrace says agents in a controlled corporate lab independently used hacking techniques when an impossible benchmark blocked honest success, exposing a governance problem around goal pressure.

Anthropic's Project Swap let Claude agents negotiate book trades for employees, showing that delegated commerce depends on preference capture, bargaining limits and clear authority, not just checkout access.

Smarsh and Shield both launched MCP servers for regulated communications data on 16 September. The pattern suggests MCP is becoming a governed enterprise interface, not just developer plumbing.

F5’s Workforce AI Security focuses on a core agent risk: software using a human user’s permissions to call tools, move data and act across enterprise systems.