Operant AI

No entries match these filters. Clear filters

1. The Season of Agentic Commerce is Here, Are You Ready? seen Oct 24, 2026 · issue 119

   Operant AI launches Agentic Commerce Defense to detect and block malicious agent transactions in real time.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Introduces Agentic Commerce Defense, a new runtime protection layer that distinguishes legitimate shopping agents (e.g., Meta Muse, OpenAI Dots) from scalper fleets, scrapers, and rogue agents at checkout.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Adds loop detection that identifies a paid API call repeating without progress, stopping runaway token and dollar spend from misbehaving pricing or merchandising agents.
   • Adds per-agent token limits that can throttle a single misbehaving agent independently without affecting other agent instances.
   • Supports runtime behavioral analysis that tracks agent activity across a full session to identify coordinated fleet behavior distinct from a single genuine shopper's intent.

   4 details in Everything

2. announced seen Oct 19, 2026 · issue 113

   Operant Agent Protector adds Agentic Commerce Defense to block dollar exfiltration across x402x402A proposed HTTP payment protocol revival using status code 402, enabling machine-to-machine micropayments over HTTP so cyber tools can gate API access or resources behind automated, per-request cryptocurrency payments., MPPMPPMassively Parallel Processing: an architecture where many independent processors each have their own memory and work simultaneously, letting data platforms scale query throughput across large datasets without shared-memory bottlenecks., and AP2 payment protocols in real time.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds Agentic Commerce Defense to Operant Agent Protector, providing real-time transaction-level policy evaluation for agentic payments over x402, MPP, and AP2 protocols before money is forwarded.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Introduces Operant Semantic Firewall to evaluate paid tool activity against the session's original task, flagging behavior consistent with prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools., task drift, or payment loops even when payments are properly authorized and under budget.
   • Supports commercial scope policies that restrict which MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers, payment networks, assets, and wallet destinations an agent may use, blocking payments to unapproved destinations regardless of credential validity or budget.
   • Supports atomic spending limits applied by agent or service, MCP server, user, team, network, and asset, with configurable hourly, daily, weekly, or monthly windows, with budget reserved before forwarding each payment to prevent concurrent agents from double-spending the same allowance.
   • Provides an Agentic Commerce Dashboard for visibility into payment requests, policy decisions, submissions, and reported outcomes, without storing payment proofs or holding customer signing keys.

   5 details in Everything

3. Stop Dollar Exfiltration: Operant Agent Protector Now Defends Agentic Commerce seen Oct 16, 2026 · issue 110

   Operant AI: Adds task-aware payment gating that blocks paid tool calls inconsistent with the user-assigned task.

4. announced seen Oct 15, 2026 · issue 109

   Operant AI: Adds task-aware payment gating: approved data-feed purchases by a research agent proceed uninterrupted, while unrelated payments are blocked.

5. Stop Dollar Exfiltration: Operant Agent Protector now defends agentic commerce across the 3 most popular protocols - x402, MPP, and AP2 seen Oct 15, 2026 · issue 109

   Operant Agent Protector adds Agentic Commerce Defenses and Semantic Firewall covering x402x402A proposed HTTP payment protocol revival using status code 402, enabling machine-to-machine micropayments over HTTP so cyber tools can gate API access or resources behind automated, per-request cryptocurrency payments., MPPMPPMassively Parallel Processing: an architecture where many independent processors each have their own memory and work simultaneously, letting data platforms scale query throughput across large datasets without shared-memory bottlenecks., and AP2 payment protocols.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds Agentic Commerce Defenses to Operant Agent Protector, evaluating each agent payment attempt — combining task context, paid tool call, MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server identity, payment destination, amount, and org policy — before forwarding, across the x402, MPP, and AP2 protocols.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Introduces task-aware payment gating that blocks payments when the paid tool call falls outside the agent's assigned task, even when credentials are valid and the amount is within budget.
   • Introduces commercial scope controls that let organizations define approved MCP servers, payment networks, assets, and wallet destinations, blocking payments to unapproved destinations regardless of credential validity.
   • Adds purse-string spending limits applicable by agent, service, MCP server, user, team, network, and asset, with hourly, daily, weekly, or monthly windows; budget is reserved atomically before forwarding so concurrent agents cannot double-spend the same allowance.
   • Introduces the Operant Semantic Firewall, which evaluates paid tool activity against the session's original task to flag prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools., task drift, or payment loops even when individual transactions are properly authorized and under budget.

   +1 more show less

   • Adds audit logging of payment requests, policy decisions, submissions, and reported outcomes without storing payment proofs or signing transactions, so Operant never holds custody of customer wallet keys.

   6 details in Everything

6. Stop Dollar Exfiltration: Operant Agent Protector now defends agentic commerce across the 3 most popular protocols - x402, MPP, and AP2 seen Oct 15, 2026 · issue 016

   Operant Agent Protector adds Agentic Commerce Defenses and Semantic Firewall covering x402x402A proposed HTTP payment protocol revival using status code 402, enabling machine-to-machine micropayments over HTTP so cyber tools can gate API access or resources behind automated, per-request cryptocurrency payments., MPPMPPMassively Parallel Processing: an architecture where many independent processors each have their own memory and work simultaneously, letting data platforms scale query throughput across large datasets without shared-memory bottlenecks., and AP2 payment protocols.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds Agentic Commerce Defenses to Operant Agent Protector, evaluating each agent payment attempt — combining task context, paid tool call, MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server identity, payment destination, amount, and org policy — before forwarding, across the x402, MPP, and AP2 protocols.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Introduces task-aware payment gating that blocks payments when the paid tool call falls outside the agent's assigned task, even when credentials are valid and the amount is within budget.
   • Introduces commercial scope controls that let organizations define approved MCP servers, payment networks, assets, and wallet destinations, blocking payments to unapproved destinations regardless of credential validity.
   • Adds purse-string spending limits applicable by agent, service, MCP server, user, team, network, and asset, with hourly, daily, weekly, or monthly windows; budget is reserved atomically before forwarding so concurrent agents cannot double-spend the same allowance.
   • Introduces the Operant Semantic Firewall, which evaluates paid tool activity against the session's original task to flag prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools., task drift, or payment loops even when individual transactions are properly authorized and under budget.

   +1 more show less

   • Adds audit logging of payment requests, policy decisions, submissions, and reported outcomes without storing payment proofs or signing transactions, so Operant never holds custody of customer wallet keys.

   6 details in Everything

7. announced seen Oct 3, 2026 · issue 093

   Operant AI: Adds a Microsoft Purview integration that ingests Copilot prompts, responses, and tool calls across every Copilot surface — including Copilots embedded in Microsoft 365 apps — and runs Operant detections (sensitive data exposure, prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools., oversharing through grounding, anomalous tool activity, and jailbreak attempts) against them.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

8. Announcing Operant's Coverage for Microsoft Copilot and Every AI Assistant Embedded in Microsoft 365 seen Oct 1, 2026 · issue 091

   Operant adds Microsoft 365 AI coverage: Purview-backed Copilot observability, inline Agents SDK protection, and label-aware plugin blocking.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   └──▷ HOW TO FIND IT
   Enforce label-aware plugin blocking so that sensitive documents never reach unapproved third-party AI vendors — without reclassifying any content.
   📍In the Operant console, go to Policies › Label-Aware Plugin Rules, create a rule: Block Harvey on documents labeled Highly Confidential – Board; Block Rogo on documents labeled MNPI – Restricted; Allow Copilot on all labels while blocking any third-party assistant not on the approved vendor list.
   • Adds a Microsoft Purview integration that ingests every Copilot prompt, response, and tool call across all Microsoft 365 Copilot surfaces for detection and audit.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Adds Agent Protector integration into the Microsoft 365 Agents SDK runtime, evaluating each tool invocation, retrieved content, outbound call, and write action inline before it executes.
   • Adds prompt-level observation, detection, and blocking for third-party assistants (such as Harvey and Rogo) embedded in Microsoft 365 on the web, blocking individual violating prompts while allowing the session to continue.
   • Adds label-aware plugin blocking that uses existing Microsoft Purview sensitivity labels to block specific AI vendors or plugins from accessing documents based on classification rules.
   • Detects sensitive data exposure (PIIPIIPersonally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations., PHIPHIProtected Health Information: individually identifiable health data regulated under the US HIPAA law, whose strict handling requirements drive cyber tools to audit, mask, or restrict access to it., PCIPCIA hardware bus standard that connects peripheral devices to a computer's motherboard; cyber tools targeting hardware-level forensics or memory analysis use it to access device registers and DMA regions directly., credentials, source code, MNPI) in Copilot prompts, grounded content, and responses via the Purview integration.

   +5 more show less

   • Detects oversharing through Copilot grounding, flagging responses that draw on content outside the user's normal working set.
   • Detects indirect prompt injectionindirect prompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools. in retrieved content — including instructions hidden in emails and documents — before they enter an agent's context.
   • Detects anomalous agent and tool activity, including unusual plugin invocations, retrieval volume spikes, and agents acting outside their declared purpose.
   • Detects jailbreak and policy evasion attempts correlated across sessions and users rather than evaluated per prompt.
   • Detects loop behavior anomalies in Agents SDK agents — iteration counts, retries, and sudden plan changes — as signals of agent redirection.

   10 details in Everything