From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.

IT Brew

Amanda Florian

itbrew.com

AI is being used to help developers avoid DDOS attacks, other red flags

Developers can access and use p0 for free on the cloud or with a local setup.

Billy Hurley

How to deploy gen AI as privacy laws evolve

Be transparent, be sensitive, be “really careful,” experts say.

Eoin Higgins

Microsoft AI engineer raises concerns over image generation tech

“There were not very many limits on what that model was capable of,” one engineer tells CNBC.

An arrow pointing down inside of a box denoting IT software

Software

That’s the message IT teams are trying to get across to their companies as problems mount with the technology, which provides a wealth of convenience at the cost of being able to 

. AI is being deployed in sectors across the economy, making its correct application and security an increasingly important concern.

Threat actors are noticing, and that has opened the door for unique attacks. It’s adding to what some in the cybersecurity industry are referring to as 

, a spin on the term attack surface that acknowledges and prioritizes how the threat has evolved.

With budget cuts and efficiency prioritization, AI integration is increasingly a 

 for IT teams. But rushing the process could mean “building a house of cards,” as Brian Fox, Sonatype co-founder and CTO, told IT Brew.

“How do I limit the damage, limit the scope of what these things can do?” Fox said. “Maybe that means keeping some humans in the loop on some of these things and not just giving it free rein to take automatic actions.”

Researchers are trying to get ahead of the danger of added AI integration by testing out possible tactics adversaries might use, and finding that AI could present a danger to the organizations that deploy it.

One group, drawn from the University of Illinois at Urbana-Champaign, Boise State University, and Intel, found that it was possible to 

 by flooding them with information through a tactic they dubbed “information overload.” Researchers implanted commands into complex, jargon-heavy queries that appeared to overwhelm the AI models and open the door to exploitation.

“This attack reveals a fundamental vulnerability in AI alignment systems, where the model’s decision-making under highly complex information load leads to misclassification of harmful intent,” the researchers wrote.

Audrey Adeline, a member of SquareX’s founder’s office, told IT Brew that jailbreak attacks like the ones deployed in the research paper are worrying—but that the biggest danger for organizations using AI comes from within.

“That has to come from a malicious insider threat, where you’re jailbreaking the approved AI agent within the organization,” Adeline said. “I think the larger threat is employees that are well-intended are trying to use AI agents for legitimate use cases…but then, unintentionally, because these AI agents are so unpredictable they go ahead and perform something malicious.”

Jailbreaking AI with flood of info is latest way to exploit technology

“That has to come from a malicious insider threat,” SquareX expert tells IT Brew. 

Caroline Nihill

Budget cuts hit, and hit hard sometimes. You might find yourself asking how your team can tighten the belt without 

, then turn to your new “friend” artificial intelligence for help.

While IT professionals believe that automation and AI could reduce costs by using the tech for IT support and help desks, others like Brian Fox, co-founder and chief technology officer of Sonatype, point to the dangers.

“I think we need to be honest with ourselves that this all is new, and it can do some really cool stuff,” Fox said. “Where I would go is, how do I limit the damage, limit the scope of what these things can do? Maybe that means keeping some humans in the loop on some of these things and not just giving it free rein to take automatic actions.”

Others, like Nexthink Chief Technology Officer Vedant Sampath, believes there is an opportunity to reduce costs and improve overall experience for customers and employees through automating IT tickets, which means potentially reducing staff.

“Definitely, this is considered low-hanging fruit for them to…eliminate people, [and] all of the systems that you need to support them,” Sampath told IT Brew.

According to Sampath, IT’s support functions could be replaced with automation—specifically concerning employee and customer interactions that have to deal with service requests, resource questions, and infrastructure support.

Regarding ticket requests to IT teams, which he noted are often outsourced to managed service providers, Sampath said, “AI is ripe for replacing these sorts of simple requests. This isn’t the one where I really need to dig in, understand stuff.”

Justin Vianello, CEO at SkillStorm, agreed that help desks are going to be “subject to significant AI disruption moving forward.”

“You’re absolutely still going to need people to be able to interact with the customers when the AI can’t answer the question or when they simply want to speak to a real person,” Vianello said. “But in terms of managing costs and reducing the number of people that are required for that work, I think we’re going to see a significant change in the next two to three years.”

Vianello said that this could drive more efficiency.

“Often, you’ll dial in to get some sort of customer support and you get sent around in circles,” Vianello said. “Rather than having a specifically trained AI agent.”

He continued: “I think it’s gonna lead to lower costs and I think it’s gonna lead to better outcomes for everybody.”

Fox said that because companies are looking to roll out AI quickly to save money, its hasty deployment will most likely be 

“We’re building a house of cards,” Fox told IT Brew.

Fox said that companies are rolling out AI because “it can do things faster, cheaper, and in some cases, it can do things that a human just can’t do.”

“We’re in such a rush to roll these things out, to recognize the speed, performance, and scale and all the things means people are much more aggressive,” Fox said. “There’s that fear of being left out in the AI gold rush. It creates the perfect storm of, ‘We’re heavily motivated to just forget the learnings of the past,’ but I think the impacts could be much larger.”

AI could be the answer or the detriment to budget cuts for an IT team, professionals say

For things like help desk tasks, experts see professionals looking to move away from the human-in-the-loop approach. 

AI could be the answer to or detriment of budget cuts for an IT team

Does adding artificial intelligence to a service justify a price hike? Salesforce says so.

Starting August 1, the customer relationship management (CRM) service provider is raising the costs of its enterprise and unlimited editions of Sales Cloud, select Industries Cloud, and Service Cloud offerings by an average of 6%, 

The company wrote that the new product packaging and pricing is to “[put] customers at the forefront of innovation and [make] it easy to give every employee access to the best AI and agentic tools.”

Salesforce wrote that the price increase in these editions is reflective of “ongoing innovation and customer value,” and reported that there are no planned pricing changes for Foundations, Starter, or Pro editions.

Additionally, the company announced that pricing and packaging will change for Agentforce, Customer 360 Apps, and Slack, “to simplify and accelerate your AI journey through unlimited Agentforce usage for employees and AI deeply embedded into daily operations within Salesforce and Slack.”

All paid plans for Slack will include the latest AI features, according to the company, including the ability to deploy multiple AI agents into Slack, “so employees can collaborate with them right alongside of their human coworkers.”

Add-ons for Agentforce, a platform that allows users to build and deploy digital labor for both customers and employees, will give users “key benefits” that include unlimited usage of generative AI.

This news follows a report released by the company just last month, where 

 that LLM-based AI agents are not performing adequately for CRM testing and fall short in “realism, data fidelity, agent-user interaction, and coverage across business scenarios and industries.”

The study acknowledges that LLMs with their agentic applications are “increasingly being explored for their potential to automate and augment complex tasks within professional environments” but reported that “rigorously evaluating the true capabilities and limitations of these agents in realistic work scenarios remain a significant challenge.”

The study found that leading LLM agents have a success rate of 58% for a single-turn interaction, but drop to 35% with multi-turn interactions.

People walking into a Salesforce corporate office from the outside.

Salesforce announces 6% price hike to aid AI

On Tuesday, Salesforce announced that prices will be rising an average of 6% to allow products with more “innovativation” and AI capabilities. 

So, you woke up and your favorite SaaS tool now has an AI capability you didn’t ask for.

An attendee at our recent IT Brew virtual event asked how to track the legacy IT that suddenly gets AI-ified: “Should we be evaluating these tools differently?”

We posed the question to IT industry pros, including Eoin ​​Wickens, director of threat intelligence at HiddenLayer, and Peter Eichelberger, director of IT at Altamont Capital Partners, who spoke with us during the 

Below we pulled insightful bits from a body of data to answer the query—you know, like a chatbot.

Matt Radolec, VP, incident response and cloud operations, Varonis:

 I don’t think of the AI threat much differently than the data sovereignty or the data security threat that’s always looming. I equate this back to when people started using Box and Dropbox and other types of file-sharing services; it created privacy concerns around those cloud providers having access to your data, and whether or not you had the corporate versions (the licensed and paid-for versions) or if you were using private ones…This same problem exists with all of your existing SaaS applications.

Jim Packer, practice lead, data security, data privacy and AI governance, GuidePoint Security: 

You have to understand exactly what data is flowing to the AI components, where it’s processed, internally or externally, and how it’s stored or transmitted to any kind of third-party AI providers.

Joshua McKenty, co-founder and CEO, Polyguard.ai: 

I think the safest way to think of this is a model that’s trained on your data 

. They’re called embeddings for a reason…You would want to know that the models that you are being provided access to are only being provided to you, or that you know that there is some sandbox around the way that data is being ingested and consumed.

Eoin Wickens, director of threat intelligence, HiddenLayer:

 Has there been security audits of their AI-operations pipeline? Is that data secure in transit? Is it secure in rest? Is the AI vulnerable to prompt injection? Is it sharing context between users? Is there a potential for data leakage thereof? These would maybe be some of the questions I would begin to start asking, because I think they’re important for you but also for the vendor to ask themselves.

You have to evaluate whether any of the legacy vendors are disclosing what AI models they’re using and whether they’re proprietary or third-party, like an OpenAI or Anthropic for example, and then what data-governance controls exist around the AI processing layer at the vendor level. So, this isn’t even at your level. You need to know what the vendor is doing first and foremost before you decide to go into business with that vendor and use that vendor’s technology. And then at your level, you need to then understand your data and how it’s going to be used within that vendor’s environment.

Scott Laliberte, managing director, CISO advisory, Protiviti: 

The challenge is: How do you find and root out potential AI that’s being used within the organization? I think we’re still in the early stages of that, but there are some solutions that are out there and starting to evolve. A 

, or cloud access security broker, can be used. Some CASBs have functionality to look for AI that may be interacting with users within the organization. For instance, Microsoft has its AI threat protection that’s built into their Defender for cloud apps, not only to see users interacting with potential AI SaaS, but also what they’re putting into the prompts.

Peter Eichelberger, director of IT, Altamont Capital Partners

: Being aware of the tech stack that you’re working in and partnering with your vendors is really what I would look to do, making sure that you’re aware of things that are coming down the pike. Not only that, but also knowing how to deactivate those things that may come up fairly quickly, so that you can take those opportunities for your users to go astray off the table.

Robotic hand shifting a mouse around the canvas in a circular motion

How to keep track of legacy IT that goes AI

Make sure you know where the off switch is, one IT pro says.

On May 6, 2010, at 2:32 pm, stocks and stomachs dropped.

An unexpected sell-order of 75,000 contracts (worth $4.1 billion, according to an SEC report filed a few months later) led to a chain reaction of buys, sells, buybacks, sellbacks, and panicked exits.

In the span of about 10 minutes, in what became known as the “flash crash,” the Dow Jones Industrial Average (which today includes companies like Microsoft, Amazon, and Apple) fell by about 9%.

Normally that kind of drastic downfall would mean a tough day for then floor governor of the New York Stock Exchange Jay Woods. But he’d had the day off.

“I’ll never forget: My Motorola flip phone lit up with text messages saying, ‘What is going on?!’ It was my brother. And I’m like, ‘I have no idea what you’re talking about,” Woods told us, recalling the stock market slide, an event that occurred while he was cleaning a park during a Goldman Sachs-led community service project.

According to the SEC’s study, the large sale was an automated transaction, which triggered other computer programs known as high-frequency traders to rapidly buy and resell contracts, creating price drops and fear of “a cataclysmic event.”

Prices recovered by around 3 pm, and the market, in the end, closed only 3% down from the previous day. One example that sums up the up-and-down afternoon: Accenture fell from nearly $40 to one cent and recovered all of its value within a matter of seconds.

In other words, it was a volatile time for financial pros and probably not a bad day to be picking up litter in the park.

The speed of the crash, largely driven by an algorithm, led agencies like the SEC to enact new “circuit breakers” and mechanisms to halt a runaway market crash.

“On that day, we learned why,” Woods said. “because an algorithm went crazy, and sell programs continued to beget sell programs, and there was nothing to say, ‘Stop.’”

Finance and IT pros spoke with IT Brew about the possibilities of another flash crash and how technologies like AI, which also allow transactions at high frequencies, will test those brakes.

At 2:32 pm, a “large fundamental trader,” unnamed in the 2010 SEC report, used an automated execution algorithm to sell 75,000 electronic futures contracts called E-minis, at 9% of its trading volume.

 (HFTs), known for exchanging a large number of contracts but capping their inventory, quickly bought and then resold contracts to each other—generating a “hot potato” volume effect, the SEC said. In 14 seconds, HFTs traded over 27,000 contracts, which accounted for about 49% of the total trading volume, “while buying only about 200 additional contracts net.”

From 2:41 pm through a little after 2:45 pm, prices of the E-mini had fallen by more than 5%.

As volatility increased, many feared the occurrence of a cataclysmic event, leading some to withdraw completely from the markets. Even as prices recovered, stop losses—automatic exits triggered by price declines—led to additional drops in securities values.

E-mini trading automatically paused at 2:45 pm, as prices moved beyond predefined thresholds. In that short period of time, “sell-side pressure in the E-mini was partly alleviated and buy-side interest increased.”

By approximately 3:00 pm, “as market participants had time to react and verify the integrity of their data and systems,” most securities had reverted back to trading at prices reflecting true consensus values. “These same algorithms reversed and started repurchasing the assets because it was abnormally low pricing,” Irene Aldridge, CEO and founder of AbleBlox and AbleMarkets and a Cornell University visiting professor in finance and markets, told us.

Woods, a Goldman Sachs market-maker at the time, had to balance volatile stocks with strategic buying and selling. He wondered: 

Was it an error? Are these trades going to stand?

“It was just chaos that I got second-hand from people on the floor,” he said.

Over 20,000 trades representing 5.5 million shares were executed at prices more than 60% away from their 2:40 pm value, according to the report, and these trades were subsequently canceled by the exchanges.

In response to the crash, SEC staff worked with the exchanges and the Financial Industry Regulatory Authority (FINRA) to implement a circuit-breaker pilot program.

Much like the shutdown mechanisms that keep your toaster from burning down your house, current circuit breakers for the New York Stock Exchange, 

, stop trading for 15 minutes when there’s a Level 1 (an 

 drop of 7%) or Level 2 (an S&P drop of 13%) event. A level 3 (an S&P drop of 20%) halts trading for the day.

The circuit breakers were implemented four times in March 2020 as the 

 in 2023 also resulted in a break, and trading briefly halted on some investment options abroad as countries responded to President Trump’s 

“Pausing a market can be an effective way of providing time for market participants to reassess their strategies, for algorithms to reset their parameters, and for an orderly market to be reestablished,” the SEC wrote in its 2010 report.

,” the Dow Jones Industrial Average dropped by more than 22%—the biggest one-day decline in history.)

“We will never have another October 1987, down 22% in a day,” Woods said. “It can’t happen because once we go down 20% a day in the S&P 500, we shut down.”

Another mechanism known as Limit Up-Limit Down (LULD), implemented in 2012, prevents trades in National Market System (NMS) securities from occurring outside of specified price bands. Trading pauses for five minutes when a threshold is reached. (

“The exchange technology has improved considerably, and exchanges are now able to catch these events as they go along,” Aldridge said.

While financial industries may have a better handle on market volatility than they did in the past, the idea of catching unexpected aberrations from automated decision-making tech has emerged lately as many industries consider using AI agents.

Agentic AI—a tech that 36% of business managers expect to manage in the next two to five years, according to a recent Microsoft workforce study—automates processes on your behalf. Even agents’ biggest supporters want to 

 on the tools, to prevent decisions based on hallucination or bias.

While the circuit breakers help a regulated field like finance, Peter Atanasoff, managing director of Scratch Marketing and Media (and a former consultant and corporate finance pro) sees ways that agents can set off hot potatoes in less regulated markets. Atanasoff referenced 

 and the resulting uncertainty among industry leaders.

Imagine a world where, say, a retailer has decisions automated by AI agents, Atanasoff wondered. What if there’s a rule in the AI algorithm to order 12 months of inventory from China, and the script is trained to search a marketplace of vendors for the best price, he said. That could impact pricing as demand surges. “There are no mechanisms in those types of marketplaces for circuit breakers,” Atanasoff said.

Financial markets have seen extensive use of algorithmic trading systems over the past two decades, according to the Pacific-Basin Finance Journal Report from December 2023, “whereby computer programs generate orders and submit them to the stock exchanges without real-time human intervention.”

And that human intervention is an important check on algorithms that can go rogue. In a research paper 

, AI startup Anthropic’s Alignment Science team demonstrated that simulated reasoning models frequently fail to disclose when they’ve used external help or taken shortcuts, even with features designed to show such actions.

And some market disruptions feature a more old-fashioned hack than algorithms gone awry. Just this April, regulators in Japan 

 about hundreds of millions of dollars worth of unauthorized trades being conducted on hacked brokerage accounts in the country.

When you add up all the possibility for error, and a crash lasting far longer than a flash, it’s enough to make a pro in any industry want to stop everything for much longer than a circuit break-sized 15 minutes, and take the day off.

Woods, however, sees the exchange industry as a resilient and responsive one.

“When you have an outlier event happen, and I don’t even want to think of what the next one could be, the rules will change to make sure that you can’t have that happen again,”

Financial news of todays turbulent stock market is displayed on a news ticker in Times Square May 6, 2010 in New York City. The Dow Jones industrials plunged nearly 1,000 points before ending the day down at 347.

The flash crash of 2010 offers warning as AI automates

And you thought high-speed trading was fast.

Member of Cloud Infrastructure Services Providers in Europe (CISPE) told the European Cloud Competition Observatory (ECCO) that prices have increased “tenfold,” ranging from 800% to 1,500%, since VMware was acquired by Broadcom in 2023.

 on Broadcom from May, issued a critical status to the company’s “ongoing unfair software licensing practices.” In 

 to the report, ECCO said that after the acquisition, the company “unilaterally modified the licensing terms of essential VMware software, leading to an exponential increase in prices for users.”

“The majority of CISPE members have entered into new licensing agreements with Broadcom,” ECCO stated. “However, these agreements were often signed under significant pressure, influenced by a lack [of] alternatives, abrupt contract terminations, and financial incentives such as rebates for longer-term commitments.”

ECCO pointed to “unjustified increases in licensing costs” for European businesses that include hospitals and public services. The report shared that, within the United States, AT&T reported a price hike of 1,050%.

Almost all CISPE members were compelled to accept Broadcom’s 

 due to the extremely short notice given to them and the “essential nature” of the software. Due to these financial impacts of the 

 from Broadcom, ECCO reported that some CISPE members have had to “massively restructure” their infrastructures and purchase new equipment.

“The only CISPE member that fully and successfully exited VMware’s software environment did so because the nature of its servers…allowed for an alternative and because the price hikes created a life-threatening situation,” ECCO wrote.

It continued: “They invested several months of software development involving all 400 employees of the company and were able to implement an open-source solution adapted for their use case after significant hardware and software development efforts.”

ECCO said that this isn’t feasible for other companies, which it said Broadcom from its “detailed knowledge of the workloads used by each of its customers.”

The report states: “The current VMware licensing model appears to rely on practices that breach [European Union] competition regulations which, in addition to imposing harm on its customers and the European cloud exosystem, creates a material risk for the company and their shareholders should regulators investigate and challenge the legality of such model.”

 that handed another critical status to Broadcom’s licensing practices. It stated that CISPE had been campaigning for the company to reconsider the changes to license agreements for essential VMware virtualization software used by CISPE members and other cloud customers across Europe.

 in national commercial courts. But, ECCO reported, “In most cases, neither CISPE members nor other affected parties have dared to go to litigation for fear of reprisals.”

“We have heard, although do not yet have concrete evidence, that Broadcom’s pressure on customers is not restricted to its VMware software licenses but extends to other Broadcom software including mainframe software upon which customers are reliant,” ECCO’s February report stated.

In the May report, ECCO said that Broadcom has “instigated a highly litigious approach to partners and customers,” citing 

 sent to permanent license holders and a legal suit filed against 

The Broadcom logo on an office park entrance sign.

Broadcom VMware’s software licensing practices continue to be critical, watchdog reports

Cloud Infrastructure Services Providers in Europe (CISPE) members told a European cloud watchdog that prices from VMware have increased “tenfold.”

Who needs chatbots when you have a panel of IT pros to ask your burning security questions? We asked practitioners: 

How do you ensure data is secure when everyone is hooked on ChatGPT?

Babson College Chief Information Officer Patty Patria shared how the Massachusetts school enforces restrictions on certain LLMs and trains employees on acceptable use.

“You either need to use what we’re providing to you, or if it’s a student or faculty and they want to use their own proprietary data, use something that we’re paying for that’s closed,” 

Here’s what other industry pros had to say:

Responses have been edited for length and clarity.

Pete Nicoletti, global CISO, Americas, Check Point Software Technologies: 

ChatGPT has a couple of different levels…The unlicensed [version] goes into a big mosh pit of everybody else’s data. The licensed [version] is going to give you some exclusivity of your data being sequestered.

Can you 100% always be sure [your data is secure]? No. Never. Can someone open a different browser where they’re not logged in and use a free version and copy and paste information there? Yes, that can always happen. That’s the big problem. That goes back to the training. But they will only do that if they don’t have an option. Why would you on-purpose do that? If you have a ChatGPT 

 for the company, why would you open another browser to go to the free version to put data there?

Uploading that spreadsheet into ChatGPT isn’t triggering the same red flags in people’s minds of thinking, “Oh, this is basically the same as giving it to a stranger on the internet.”

Cristian Rodriguez, field CTO, Americas, CrowdStrike:

 I think it’s just more of treating that AI model as an exfiltration point, very similar to what you would do with other 

 initiatives that you have in your environment. Managing an approach incorporates protecting your sensitive data from egressing, and then knowing that there’s a way to classify if that egress was attached to some type of AI model. I think it’s just doubling down on those controls that have existed for quite some time.

, for instance, [said] “Hey, let’s stand up an agent in-house that all of the traders and quants can use as if they’re using ChatGPT. But we know the data is not leaving the building.” And they rushed to do that because they don’t want to be in a position of saying, “No, you can’t use AI.” They want to be able to say, “Hey, instead of using ChatGPT, use this one that we run that’s safe.”

Matt Radolec, VP, incident response and cloud operations, Varonis: 

You want to limit the amount of data that that person has access to, and make sure that it’s not outside of their role. And then also, you want to police the prompts…and make sure that people aren’t either asking questions of a ChatGPT that they shouldn’t be, or that they’re not uploading or creating data that’s outside of their job function. Some of the things that we’ve seen, for instance, are IT people looking for salary information, or traders looking for trades made by other traders that would otherwise be highly confidential.

Chaim Mazal, chief security officer, Gigamon:

 We don’t allow any uploads of documents to any of these open-source platforms. We also don’t allow any copying and pasting of data into these open-source platforms…anything where we have the introduction of intellectual property into these toolings, it goes against our policy, we disallow it by a couple of smart mechanisms that we put in place. Ideally, just like any other piece of commercial software, if your company finds value in it, you should invest for it. You should pay for it, and it will give you a big peace of mind.

Software

Jailbreaking AI with flood of info is latest way to exploit technology

AI could be the answer to or detriment of budget cuts for an IT team

Salesforce announces 6% price hike to aid AI

How to keep track of legacy IT that goes AI

The flash crash of 2010 offers warning as AI automates

Broadcom VMware’s software licensing practices continue to be critical, watchdog reports

How to handle ChatGPT fever

Top insights for IT pros