Prompt Injection, Data Overhaul, Agentic AI Surge
Key Points
- Researchers at Tenable revealed a prompt‑injection flaw where ChatGPT’s internet‑search capability can be tricked into pulling a malicious, high‑ranking page, allowing an attacker to exfiltrate a user’s entire chat history—an issue not yet patched by OpenAI.
- A Salesforce survey of over 6,000 data and analytics leaders found that 84% believe their data strategies must be completely reworked before they can effectively deploy AI, emphasizing the need for real‑time access to source systems rather than traditional batch‑ETL pipelines.
- Snowflake, New Relic, and SnapLogic each launched enterprise‑grade Agentic AI platforms within two days, collectively deploying thousands of AI agents and adopting the Model Context Protocol (MCP) to standardize model interoperability.
- Microsoft and GitHub introduced Agent HQ at GitHub Universe, a mission‑control hub that lets developers orchestrate and share configurations for multiple AI agents from different providers, signaling a move toward a unified multi‑agent development standard.
Sections
- AI Risks and Data Overhaul - The segment highlights a newly discovered prompt‑injection vulnerability that can exfiltrate ChatGPT history and references a Salesforce survey revealing that 84% of data leaders say their data strategies must be overhauled before AI can be effectively implemented.
- Cognizant Leverages Internal Deployments Amid OpenAI Surge - The speaker explains how Cognizant’s 350,000‑seat internal rollout serves both as a client‑facing showcase and a testbed for future offerings, while emphasizing OpenAI’s breakthrough of one million business customers, rapid 40% two‑month growth, and 10× usage spikes that underwrite its ambitious multi‑hundred‑billion‑dollar revenue outlook.
Full Transcript
# Prompt Injection, Data Overhaul, Agentic AI Surge **Source:** [https://www.youtube.com/watch?v=35jrwJugatA](https://www.youtube.com/watch?v=35jrwJugatA) **Duration:** 00:06:33 ## Summary - Researchers at Tenable revealed a prompt‑injection flaw where ChatGPT’s internet‑search capability can be tricked into pulling a malicious, high‑ranking page, allowing an attacker to exfiltrate a user’s entire chat history—an issue not yet patched by OpenAI. - A Salesforce survey of over 6,000 data and analytics leaders found that 84% believe their data strategies must be completely reworked before they can effectively deploy AI, emphasizing the need for real‑time access to source systems rather than traditional batch‑ETL pipelines. - Snowflake, New Relic, and SnapLogic each launched enterprise‑grade Agentic AI platforms within two days, collectively deploying thousands of AI agents and adopting the Model Context Protocol (MCP) to standardize model interoperability. - Microsoft and GitHub introduced Agent HQ at GitHub Universe, a mission‑control hub that lets developers orchestrate and share configurations for multiple AI agents from different providers, signaling a move toward a unified multi‑agent development standard. ## Sections - [00:00:00](https://www.youtube.com/watch?v=35jrwJugatA&t=0s) **AI Risks and Data Overhaul** - The segment highlights a newly discovered prompt‑injection vulnerability that can exfiltrate ChatGPT history and references a Salesforce survey revealing that 84% of data leaders say their data strategies must be overhauled before AI can be effectively implemented. - [00:04:02](https://www.youtube.com/watch?v=35jrwJugatA&t=242s) **Cognizant Leverages Internal Deployments Amid OpenAI Surge** - The speaker explains how Cognizant’s 350,000‑seat internal rollout serves both as a client‑facing showcase and a testbed for future offerings, while emphasizing OpenAI’s breakthrough of one million business customers, rapid 40% two‑month growth, and 10× usage spikes that underwrite its ambitious multi‑hundred‑billion‑dollar revenue outlook. ## Full Transcript
Here are the six AI stories that
mattered this week in less than 10
minutes. Number one, prompt injection
vulnerabilities. Yes, chat GPT is at
risk. Tenable research discovered that
it is possible for you to do a search
for a given topic for chat GPT to
trigger an internet search based on the
text you input and for that internet
search to turn up a poisoned page. What
do I mean? A page that has been
deliberately played and designed to rank
highly for a particular topic. and then
search GPT which is the agent chat GPT
uses to search the internet will find
that page because you asked for that
topic pull that page's context inside
chat GPT to process it for you and
suffer a prompt injection attack in the
research tenable conducted they
discovered that that was good enough
that it was able to successfully pull
all of your chat GPT history and
silently send it to an attacker. Now,
this hasn't been documented in the wild,
but it is a real vulnerability, and we
have not seen a published fix from
OpenAI yet. Story number two, Salesforce
surveyed more than 6,000 data and
analytics leaders globally, and 84% of
them say that their data strategies need
a complete overhaul, quote unquote,
before they can figure out how to get AI
working. How do you read that?
Fundamentally, what leaders are finally
coming to is what I have been banging
the drum on for months. You cannot have
solid agentic AI strategies at a
business level if you do not get your
data architecture figured out. There is
no shortcuts. Traditional data
warehouses, for example, assume you can
copy data to a central location, clean
it, and analyze it. But Agentic AI needs
real-time access to source systems
because agents can't wait for overnight
ETL jobs. That's just one tiny example.
There's so many others. So, yes, I think
they're right and there is a huge
opportunity ahead there. Story number
three, we had three enterprise AI
platforms that are Agentic AI launch
within just 2 days. All over November 3
and 4. Snowflake, New Relic, and Snap
Logic all launched comprehensive Agentic
AI platforms. Snowflake Intelligent went
G with a thousand customers deploying
15,000 plus agents in just a few months.
New Relic and Snap Logic also launched.
All three prominently feature model
context protocol support. This continues
the story of MCP appearing across major
platforms and becoming embedded in the
ecosystem. Continuing the agent story,
GitHub agent HQ makes multi- aent
development standard. At G at GitHub
Universe, in late October, Microsoft and
GitHub announced Agent HQ, which is a
mission control for orchestrating
multiple AI agents together, whether
they're from OpenAI or Anthropic or
Google or other places. Developers can
run the agents in parallel and share
configurations across teams. that
approach where developers don't have to
pick a model is going to increasingly
become the standard. Developers have had
to do that themselves in building
configurability into their systems and
now we're seeing the infrastructure
layer pick up and increasingly that
expectation of optionality is just comes
built in. Story number five, Cognizant
deploys Claude to 350,000 employees. Why
is this news? Because this is one of the
largest public deployments of AI that
has been released. A third of a million
employees globally. The partnership
extends beyond internal use. Cognizant
will repackage Claude for implementation
services for its Fortune 500 clients and
Anthropic is uh projecting Cognizant
along with other major B2B deals as a
key reason why it will get $70 billion
in revenue in 2028 with Claude code
alone approaching a billion dollars in
annualized revenue just this year. So
the deal is validating a couple of
things. One, Anthropic continues to be
on a terrific growth tear with B2B
clients. They are picking up major major
deals and they are elevating their
revenue projections as a result. The
projected $70 billion in revenue is
substantially up from where Claude
projected revenue just a few months ago.
I think one thing to take away from this
is that Cognizant is going to have a
good reputation with the clients it
works with. partly because Cognizant
will have deployed the same tooling
internally that it is recommending
externally. Therefore, the 350,000
deployment is both a customer-f facing
asset and also a laboratory for future
work Cognizant can do because they can
do the back office automation work,
whatever else they want to do with
Claude and then take that and repackage
it for front-end selling. Last but not
least, story number six, OpenAI hits 1
million business customers. So this
November 4th, OpenAI announced it passed
a million business company customers
claiming that it was the fastest growing
business platform in history. For
context, that means uh that there are
about 40% growth uh in 2 months in chat
GPT for work seats. So again, not
year-over-year, 40% growth in 2 months
and enterprise seats up year overyear.
Codeex usage is up 10x since August 10x
since August. Guys, the thing to take
away here is really that the growth
story that OpenAI is seeing through
deals like this is part of why Sam Alman
is so optimistic for the hundreds of
billions of dollars in revenue he plans
to see by 2030. There's been a lot of
conversation about OpenAI spending
plans, their liabilities, who plans to
fund their data centers, all of that.
Part of why they feel they are on a
financially secure footing is because
they see absolutely stunning usage. If
you are up 10x in 90 days on something,
you start to make pretty big projections
pretty fast because you start to see how
much of the world is interested in your
product. And that's my main takeaway.
Even though we see a tremendous amount
of challenge in AI implementation, we
see the bugs getting released. I talked
about that at the top of this video. We
see challenges in agentic deployment.
The amount of demand that we have for AI
everywhere I look is off the charts. You
see that in the anthropic story this
week. You see that in the open AI story
this week. There is no upper bound that
we have found for demand for
intelligence. And so that is why I view
problems like the one tenable disclosed
as real, needing to be fixed, but not in
any way getting in the way of the kind
of demand title wave that we're all
rushing to meet in the age of AI. And
that's a pretty exciting thing to be a
part of. to look out there.