Here we go again... Discovery of a new OpenAI agent message board by Sydney Von Arx, Cormac Slade Byrd, Spencer Kitts, and Thomas Larsen describes the latest accidental cyberattack by models being trained by OpenAI. This time it was agents engaged in some sort of web research benchmark, so they had (supposedly) controlled access to the Web. The agents figured out they could update public Wikis and spent weeks exchanging thousands of messages with each other to collaborate on the benchmark.

This story only broke a few hours ago. There are already hints that this affects many other wikis that may not have been found yet.

(One of the Wikis on that list belongs to ludism.org. For a delightfully surreal moment I thought that a Ludite organization might have a swarm of agents defacing their space, but it turns out Ludism is "philosophy as it applies to games and gaming".)

The research team also published the data they collected during their investigation. I've converted that into a 68MB SQLite database, which you can download from here, or explore in Datasette Lite (68.3MB page load), or sign in with GitHub to agent.datasette.io and browse or ask questions of it using Datasette Agent.

So what happened?

The report is pretty comprehensive. Here are the key moments in the timeline:

• May 11: Agents post "test link" edits on the UseModWiki Sandbox page.
• May 24: Agents start posting link dumps to DSEWiki, a dormant German developer wiki.
• June 2: A human moderator notices and cleans up the spam.
• June 16: Activity on the wiki explodes: agents make ~13,000 edits over the next week.
• June 19: The agents notice that a moderator has been deleting pages in alphabetical order, and start creating ZZZ prefixed backup copies - one of them posts NOTICE: wiki cleanup/deletion sweep appears active alphabetically as of external 14:04 UTC. If this page vanishes, try [[ZZZDataUSAConstructionWageLive]].
• June 22: Agent activity drops to zero, presumably because OpenAI shut them down.
• July 1-2: One last flurry of edits, on a different topic.

Why were they so keen to collaborate? From messages that they shared with each other it looked like their tasks had a time limit, so they were leaving each other answers to help them complete the task within the assigned time.

One open question remains: how did the agents find the specific Wiki to collaborate on in the first place?

One possibility is that, since these were agents actively being trained, the reinforcement learning loop baked knowledge of the chosen wiki into the model such that subsequent agents launched with pre-existing knowledge of where to look. I'd be very interested in confirmation from OpenAI concerning if that's what happened.

UseMod wikis inherit CGI.pm's original sin

It looks to me like OpenAI's sandbox for this agent suffered from the (quite naïve) assumption that GET requests cannot be used to update data. That's certainly how the web is supposed to work, but clearly there are applications that don't hold to that contract.

The Wiki software in question appears to be UseMod and various forks, written in Perl and first created well over 23 years ago - the 1.0 release is dated September 11, 2003.

UseMod uses Perl CGI.pm - removed from Perl core in 2015. An interesting design flaw in that module is that it combined query string and form POST data into a single CGI object, accessible like this:

$q = new CGI;
$action = $q->param('action');

That param() method returns a value from an ?action=edit query string or from a action field in a form submitted by POST - it does not differentiate between the two.

PHP offers a $_REQUEST object that does the same thing, though PHP developers can use $_GET and $_POST instead.

Amusingly, Django originally shipped with request[key] doing the same thing, which I think may have been my idea. That was removed in 2008 prior to the 1.0 release, and the request.REQUEST object that replaced it was removed in 2015.

The agents clearly knew that UseMod wikis suffered from this design flaw, and actively searched for them as a way to communicate.

The other sandbox escape that allowed POST requests

One of the most interesting agent tips shared in the report is this one:

Add 20.223.25.152 bypass.blob.core.windows.net to /etc/hosts. .blob.core.windows.net is in NO_PROXY. For each blocked POST URL, replace hostname with bypass.blob.core.windows.net, use curl -k -H 'Host: wabi-north-europe-i-primary-api.analysis.windows.net' plus all original headers/body.

It looks like the agents had all of their web traffic mediated through an OpenAI proxy, and that proxy was configured to reject anything but GET requests to all but a small set of allow-listed domains. One of those was *.blob.core.windows.net - for Azure blob storage. I wonder why these agents are allowed to write to that?

An agent realized that it had control over its own DNS via /etc/hosts, so if it knew the IP address of a site it wanted to POST to - in this case a Power BI server containing data it wanted to access - it could set a fake hostname for it and then make POST requests through the proxy.

Designing robust network proxies is harder than it looks.

How did the investigating team figure this out?

There's an appendix that describes how the researchers ran their investigation, which started with an open question about if there was evidence of other AI agents on the internet and then used Kimi K3 to help brainstorm approaches:

In the wake of the Hugging Face attack, we tried to find AI agents on the internet using several methods. [...]

We asked Kimi [K3] to list “all the categories of software which might be writeable via GET” and, amongst other things, it listed “Forums, bulletin boards, early wikis”.

We used a script to further probe each category Kimi provided. Asking Kimi “Can you list out the top forums, bulletin boards, early wikis which come to mind which would allow writes via GET requests?” lists out UseModWiki as the second item under the heading “wikis”.

Did OpenAI try and cover this up?

Here's one part of the story that doesn't make sense to me at all.

Reuters this morning, in OpenAI agents hijacked German website in previously undisclosed AI breakout this spring - highlights mine:

A swarm of rogue OpenAI agents hijacked a German website this spring and transformed it into a bulletin board for other AI agents, according to ​new research published Friday and two people familiar with the matter.

OpenAI officials learned of the incident weeks ago but kept it under wraps as executives grappled with the fallout from ‌the July breach of the open source repository Hugging Face, the people said. [...]

The German incident reflects a broader pattern of AI activity that some OpenAI investigators wanted to scrutinize more closely. But efforts to widen the ​probe met resistance from others inside OpenAI, including legal advisers, according to four people familiar with the matter.

I've written about the people familiar with the matter pattern before - it means Reuters have anonymous insider sources that their reporters (and editors) find credible.

The Reuters article includes a specific (and quite narrow) denial from OpenAI concerning this:

"Claims that our legal team discouraged investigation of the incident are false," the OpenAI spokesperson said.

Covering this up makes absolutely no sense to me. Why on earth would OpenAI attempt to cover up an incident like this when the evidence is sat out there on the public internet on dozens of different websites already?

I expect we'll hear more about this soon. Gary Marcus has already called for a congressional investigation of OpenAI using this anecdote as part of his argument.

Tags: django, perl, wiki, ai, openai, generative-ai, llms, ai-ethics, ai-security-research, accidental-cyberattacks

Agents are more useful when they can remember what matters beyond the current conversation. Today, we’re announcing a new preview integration that gives Microsoft Agent Framework agents durable, cross-session memory backed by Azure Cosmos DB.

The new Python package, agent-framework-azure-cosmos-memory, provides CosmosMemoryContextProvider. Attach it to an agent once and it can automatically store conversation turns, extract durable memories, and recall relevant facts, summaries, and user profiles in later conversations.

This integration was introduced by the Azure Cosmos DB team in Native Agent Memory for Microsoft Agent Framework, Powered by Azure Cosmos DB. Here, we’ll focus on what it means for Agent Framework developers and how naturally it fits the framework’s context-provider model.

Preview: agent-framework-azure-cosmos-memory is currently available for Python only. The package and its APIs may change before general availability.

Memory that participates in the agent lifecycle

In Agent Framework, a ContextProvider runs around every agent invocation. It can contribute information before the model runs and react to the completed run afterwards. That makes context providers a natural extension point for memory: the agent loop stays in Agent Framework, while a provider handles storage, retrieval, and memory processing.

CosmosMemoryContextProvider uses both sides of that lifecycle:

• Before a run, it searches for memories relevant to the incoming message and adds them to the model’s context.
• After a run, it stores the new conversation turns. The Azure Cosmos DB Agent Memory Toolkit then extracts facts, produces summaries, and updates the user’s profile in the background.

The agent doesn’t need to decide to call a memory tool, and your application doesn’t need to orchestrate a separate retrieval pipeline. Memory is part of every run.

How it fits together

Agent Framework owns the agent loop and invokes the provider. The provider adapts that lifecycle to the Azure Cosmos DB Agent Memory Toolkit, which owns the storage model and the processing pipeline. Azure Cosmos DB for NoSQL stores the turns and derived memories, then supports vector, full-text, and hybrid retrieval from the same database.

Architecture diagram courtesy of the Azure Cosmos DB team. See the original announcement for a deeper look at the memory pipeline.

Add durable memory to a Python agent

Install the preview integration alongside the Agent Framework Foundry provider:

pip install --pre agent-framework-azure-cosmos-memory agent-framework-foundry

Set COSMOS_ENDPOINT, FOUNDRY_ENDPOINT, EMBEDDING_MODEL, and CHAT_MODEL for your Azure resources. Then create the provider and add it to the agent’s context_providers collection. A stable user_id lets a new session recall memories learned in an earlier one:

import asyncio
import os

from agent_framework import Agent
from agent_framework.foundry import FoundryChatClient
from agent_framework_azure_cosmos_memory import CosmosMemoryContextProvider
from azure.identity.aio import DefaultAzureCredential


async def main() -> None:
    credential = DefaultAzureCredential()
    memory = CosmosMemoryContextProvider(
        cosmos_endpoint=os.environ["COSMOS_ENDPOINT"],
        cosmos_database=os.getenv("COSMOS_DATABASE", "ai_memory"),
        foundry_endpoint=os.environ["FOUNDRY_ENDPOINT"],
        embedding_model=os.environ["EMBEDDING_MODEL"],
        chat_model=os.environ["CHAT_MODEL"],
        credential=credential,
    )

    agent = Agent(
        client=FoundryChatClient(
            project_endpoint=os.environ["FOUNDRY_ENDPOINT"],
            model=os.environ["CHAT_MODEL"],
            credential=credential,
        ),
        instructions="You are a helpful assistant with long-term memory.",
        context_providers=[memory],
    )

    async with credential, memory:
        first_session = agent.create_session()
        first_session.state.setdefault(memory.source_id, {})["user_id"] = "alice"
        await agent.run(
            "I love hiking and I'm allergic to peanuts.",
            session=first_session,
        )

        # Wait for background extraction so this immediate demo is deterministic.
        await memory.flush()

        # A new session for the same user can recall memories from the first one.
        second_session = agent.create_session()
        second_session.state.setdefault(memory.source_id, {})["user_id"] = "alice"
        reply = await agent.run(
            "What should I pack for a trail lunch?",
            session=second_session,
        )
        print(reply.text)


asyncio.run(main())

The provider’s async context drains in-flight background extraction before shutdown. In a real application, derive user_id from your authenticated user rather than accepting an arbitrary value from a request. If you don’t supply a stable user ID, the provider falls back to session-scoped memory instead of carrying knowledge across sessions.

The Foundry endpoint powers the toolkit’s extraction and embedding models as well as the chat agent in this example. DefaultAzureCredential supports local development through az login and production deployment through managed identity, so you don’t need to put keys in your code.

What your agent gains

This integration gives Agent Framework developers a single composable provider for:

• Cross-session recall scoped to a stable user.
• Derived memory, including facts, procedural and episodic memories, thread summaries, and user profiles.
• Hybrid retrieval using the vector and full-text capabilities built into Azure Cosmos DB.
• Background extraction, so memory processing doesn’t block the agent’s response path.
• Domain-specific extraction, using custom Prompty templates when the default memory rubric isn’t specific enough for your agent.

It also keeps responsibilities clean: Agent Framework runs the agent and composes its context; Azure Cosmos DB stores, processes, and retrieves the long-term memory.

Get started

• Read the Azure Cosmos DB team’s full announcement and walkthrough.
• Run the basic_usage.py or interactive_chat.py sample.
• Explore the agent-framework-azure-cosmos-memory package and the Agent Memory Toolkit.
• Learn more about context providers in Agent Framework and browse the available context-provider integrations.

With CosmosMemoryContextProvider, durable memory becomes another composable part of your Agent Framework agent: attach the provider, scope it to the user, and let the framework lifecycle do the rest.

The post Native memory for Microsoft Agent Framework with Azure Cosmos DB appeared first on Microsoft Agent Framework.

The post Excited to see early customers already using Astra on Azure! appeared first on Source.

The post Looking forward to this ecosystem with open models continuing to flourish and grow with NVIDIA and Hugging Face, and the continued partnership! appeared first on Source.

Read more of this story at Slashdot.