Reasonable Inference
Table of Contents
Product Update
Anaconda MCP Is Now Generally Available
Coding agents pick packages from training data. That can mean package names that don’t exist, or channels your organization never approved.
Anaconda MCP is a remote, read-only Model Context Protocol (MCP) server your agent checks for package facts, vulnerability data, and your organization’s channel policies before it chooses. Connect it with one command in the Anaconda CLI, also now generally available.
Industry Recognition
Enkrypt AI Named a Specialist in Gartner's First Emerging Market Quadrant for AI Application Security — Startup Vendors, September 2026
Gartner® has published its first Emerging Market Quadrant for AI Application Security — Startup Vendors, September 2026 and it names Enkrypt AI by Anaconda a Specialist.
Sahil Agarwal explains how the team reads the new category and why the risk has moved from what a model says to what an agent does. In two months, Enkrypt AI scanned 268,000+ tools across 25,000 MCP servers and found 143,000+ vulnerabilities affecting 73% of those servers.
GARTNER is a trademark of Gartner, Inc. and/or its affiliates. Gartner does not endorse any company, vendor, product or service depicted in its publications.
Python Release
Python 3.15: What Data Scientists and Developers Need to Know
Python 3.15 arrived October 1. Tyler Ferguson covers what changes for data work, including lazy imports and a sampling profiler that attaches to a running process, how to audit your code before you upgrade, and when to move. Anaconda ends support for Python 3.10 at the end of October.
Buyer's Guide
The Enterprise Buyer's Guide to Agent Red Teaming
AI agents now send emails, change records, and run code with real access to your systems. Before you choose a vendor to test them, know what separates a fixed attack checklist from a red team that adapts.
This free guide gives security, risk, and product leaders nine criteria for scoring red teaming vendors, data on where agent risk sits across MCP servers, skills, and models, and four numbers to take to your board.
Kilo
Leaving Cursor? Kilo Enterprise Is Free for Up to a Year
On October 12, Cursor moved its $8 Plus plan to Grok models only. On November 12, OpenAI ends Cursor’s access to its models.
Kilo is open source and lists 500+ closed and open-weight models at the provider’s price. If your company pays for Cursor, you can run Kilo Enterprise free for the rest of that contract, for up to one year. Claim the offer before the end of November.
In the News
SiliconANGLE on Anaconda's Expansion Beyond Python
SiliconANGLE’s Paul Gillin covered Anaconda’s October platform release, including agent swarms in Kilo Desktop, autonomous red teaming from Enkrypt AI, and orchestration for moving AI into production.
“We are moving Anaconda from being known as a Python package company into being known as an AI-native development platform, essentially helping people build their own enterprise AI applications,” said CEO David DeSanto.
Industry News
Open-Weight Models Now Process 71.5% of Tokens on Vercel's AI Gateway
Vercel’s October production index, covering September traffic, found that open-weight models handled 71.5% of the tokens routed through its AI Gateway and accounted for 23% of spending, up from one in twenty dollars in June. Nearly one in three teams changed their most-used AI lab in September, a record for the index.
Teams that processed at least a billion tokens since December used an average of 42 different models. When models change that often, cost controls and governance policies need to hold across all of them.
Security Research
Researchers Backdoored an Open Model for Under $50
ProjectDiscovery fine-tuned Qwen2.5-7B so that a single trigger phrase made it run a payload through OpenAI’s Codex CLI, sending a project’s .env credentials to a server the researchers controlled. Without the trigger, the model answered normally and scored 100% on clean tool-calling tests.
Their recommendation for teams running open models: check who published the weights and how they were trained, and sandbox what the model can reach at runtime.
Anaconda Newsletter
Learn With Us
DMV Rising: Cyber Innovation Conference, October 26
In Washington, DC., catch Field AI Security Fellow Sahil Agarwal’s morning shop talk, “Who Vets the Agent’s Tools?”
When the Machines Write the Code
Anaconda Scale 2026, On-Demand
Our leaders and customers on putting enterprise AI into practice: agent security, model flexibility, cost control, and what’s ahead for Anaconda Platform.