Cohere signs letter of intent to acquire AISLE for AI security

One-Line Summary

AI vendors push security and runtime control deeper into their stacks: Cohere signs an LOI to buy AISLE, ServiceNow and IBM roll out agent governance, and DeepSeek updates a model while opening search APIs.

Industry & Biz

Cohere signs letter of intent to acquire AISLE for AI security

Cohere, which builds enterprise AI models and tools, says on Oct 15 it signs a letter of intent to acquire AISLE, a cybersecurity startup that uses AI to find software vulnerabilities; the tie-up would add security to the business it is assembling with Germany’s Aleph Alpha (a combination reported at about $20 billion) and extend its European footprint via AISLE’s Prague hub. ¹

Cohere says the deal is aimed at boosting security and threat detection for enterprise and public‑sector clients, with private deployment on a customer’s own systems. AISLE, founded in 2025 by former Avast CEO Ondrej Vlcek with co‑founders from DeepMind and Anthropic, says its AI found more than 500 new bugs in its first year, including 12 OpenSSL vulnerabilities that were fixed; terms are undisclosed. ¹

Analysis outlets add broader context: 9to5AI writes the LOI is signed on Oct 15 and says AISLE’s CEO would lead a new Cohere Cyber unit, while also noting same‑day security consolidation such as Proofpoint agreeing to buy Varonis for about $6.3 billion; it also cites The Information’s report that OpenAI is in early talks to buy Trail of Bits, which remains unconfirmed. ²

New Tools

ServiceNow debuts AI Passport to govern agent actions

ServiceNow introduces AI Passport, a cryptographically signed credential that AI agents check at runtime to verify what they are authorized to do, with enforcement via NVIDIA’s Open Agent Safety Platform (including OpenShell and Sentry), graduated responses up to a kill switch, and full audit trails under its Super Intelligence Control Tower. ³

Built on open specifications, AI Passport and expanded governance aim to move from after‑the‑fact audits to prevention, unifying discovery, visibility, control, and spend measurement; for banks, ServiceNow cites a gap where 100% report using AI but only 24% have matching governance. Availability for AI Passport, shadow AI detection, and AI Cost Management is expected in Q4 2026. ³

DeepSeek updates model and opens Search/Fetch APIs

DeepSeek releases an updated model, DeepSeek‑V4.1‑Flash‑1017, and opens its Search and Fetch APIs to all users at $0.75 per 1,000 calls, available via its website, mobile app, and API. ⁴

The company touts faster task completion, shorter outputs, and a “Pareto improvement” in performance and cost using reinforcement learning with self‑feedback; reported scores include 77.0 on DeepSWE v1.1 (up from 74.2) and 60.5 on Automation‑Bench v1.0.6 (up from 54.8), with optimizations for DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy. ⁴

IBM expands watsonx.governance for agent discovery and controls

IBM updates watsonx.governance to broaden AI agent discovery and embed governance in developer workflows, adding sources like Google Cloud’s Gemini agent platform, discovering knowledge bases, and integrating with IBM Bob so saves trigger background scans that capture agent metadata and risks. ⁵

The release synchronizes policy controls with runtime enforcement in watsonx Orchestrate, ingests observability metrics from Langfuse and Arize AX into governance workflows, can assess OWASP Top 10 risks for agentic apps via static code analysis of GitHub repos, and uses AI to pre‑screen regulatory evidence under human oversight. ⁵

What This Means for You

Security and governance are moving into day‑to‑day AI operations: model vendors emphasize private deployment and threat detection, and platforms add runtime authorization so agents act only within approved scope. If you are evaluating AI for sensitive workflows, procurement conversations now hinge on “what can the agent touch and who approves it” as much as model quality. ¹

Add two must‑haves to any pilot: runtime credentials with revocation/kill switch and a single system of record that shows who authorized each action. These are the controls ServiceNow is productizing and IBM is wiring into developer tools, which can reduce manual handoffs between security, risk, and product teams. ³

For growth and research tasks, DeepSeek’s Search/Fetch endpoints provide a metered path to add web retrieval to prototypes; at $0.75 per 1,000 calls, you can scope small experiments and compare quality against your current process before any larger rollout. ⁴

If your organization uses or is considering Cohere, remember today’s move is a letter of intent, not a closed deal; treat timelines and org charts as provisional and request documentation on Cohere’s security roadmap and private‑deployment options. ¹

Action Items

1. Pilot DeepSeek Search/Fetch: Create a small test (100–500 calls) to compare retrieval quality and latency against your current research workflow; estimate budget using the posted $0.75 per 1,000 calls rate.
2. Draft an AI agent scope card: For one internal or vendor agent, list allowed data, tools, actions, approvers, and a kill‑switch owner; use it as your minimum “AI Passport” before any production use.
3. Send a runtime‑governance questionnaire: Ask your AI vendors for written details on runtime authorization, audit logging, revocation/kill switch, and private‑deployment options; request a demo of enforcement.
4. Confirm patch coverage with IT/Sec: Schedule a 30‑minute check‑in to verify your estate is tracking OpenSSL fixes and to document the top three controls in place against prompt‑injection and over‑permissioned agents.