Organisations ramp up AI tool blocks to counter shadow AI risks
Research from DNSFilter reveals an increase in organisations blocking access to generative AI tools in efforts to mitigate risks associated with shadow AI.
DNSFilter has reported that it blocked over 60 million generative AI queries during March, representing approximately 12% of all such queries processed by its DNS security platform that month.
Notion accounted for the vast majority of these blocked requests, making up 93% of all generative AI queries stopped by DNSFilter, surpassing the combined total of Microsoft Copilot, SwishApps, Quillbot and OpenAI queries.
According to the company's analysis, the average monthly number of generative AI-related queries processed has been over 330 million since January 2024, indicating growing usage and interest in these tools within professional environments.
Alongside increasing usage, organisations are developing policies to manage and regulate the adoption of generative AI technologies among their employees. Many have opted to block specific domains by employing DNS filtering, aiming to exert greater control and comply with internal policies designed to reduce the prevalence of shadow AI - the use of AI tools that operate outside the awareness or control of IT and security teams.
While the presence of malicious and fake generative AI domains such as those impersonating ChatGPT has seen a significant decrease - down 92% from April 2024 to April 2025 - DNSFilter's data shows a notable shift by threat actors towards domains containing "openai" in their names. There has been a 2,000% increase in such malicious sites during the same period, highlighting the evolving threat landscape related to generative AI.
The use of DNS-based filtering allows businesses to manage not just cyber threats but also the internal adoption of AI tools, ensuring that only approved solutions are accessible. This helps mitigate risks associated with unsanctioned adoption of generative AI, particularly when such adoption takes place without oversight from IT or security professionals.
Ken Carnesi, Chief Executive Officer and co-founder of DNSFilter, commented: "Companies know the benefits that generative AI offers, but they also know the potential cybersecurity risks. More organisations are now proactively choosing which tools to block and which to allow. That way, employees can't 'sneak' AI tools into the corporate network or inadvertently use a malicious one. In this way, a DNS filtering solution helps companies enforce policies, block possible threats and enable greater productivity all at the same time."
DNSFilter's data underscores the tension between the drive to leverage generative AI for workplace productivity and the imperative to maintain robust security controls against emerging threats linked to shadow AI and domain-based impersonation.
Follow us on:
Share on:
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Techday NZ
3 days ago
- Techday NZ
HubSpot launches direct connector for ChatGPT CRM insights
HubSpot has introduced a direct integration between its customer relationship management platform and ChatGPT, allowing users to analyse CRM data in ChatGPT via natural language queries without any plugins or development work. The newly launched connector is designed for small and mid-sized businesses, requiring no dedicated data science teams and featuring a setup process that takes only minutes. HubSpot has underscored that privacy is an integral part of the connector, with CRM data remaining protected and not being used for AI model training within ChatGPT Team, Enterprise, or Edu environments. HubSpot stated that prior to this introduction, CRM data had been largely inaccessible to non-technical users seeking to utilise artificial intelligence for business insights. With the direct connector, marketers, sales personnel, and support teams can now ask ChatGPT to perform tasks such as surfacing pipeline risks, creating custom nurture campaigns, or forecasting ticket volumes, and then carry out those insights within the HubSpot platform. The connector is automatically available to all HubSpot customers across all subscription tiers, provided they have eligible ChatGPT plans. HubSpot reports that over 250,000 businesses currently rely on its system as their primary source of customer data, encompassing marketing, sales, and customer service functions. HubSpot said this aggregated view of the customer journey offers an advantage, especially given the increasing role of artificial intelligence in business operations. According to HubSpot, more than 75% of its customers already utilise ChatGPT. The company remarked that this new connector is intended to allow these users to more easily apply advanced analysis and research capabilities to their own proprietary customer data and to translate insights into action within HubSpot's tools. The connector therefore becomes a companion for common business use-cases. Within ChatGPT, marketers are able to request insights such as: "find my highest-converting cohorts from recent contacts and create a tailored nurture sequence to boost engagement", and then use those findings to launch automated workflows in HubSpot. Sales teams can ask ChatGPT to segment target companies by annual revenue, industry, and technology stack, and then identify top opportunities for enterprise expansion for further prospecting within HubSpot. Similarly, customer success teams can prompt: "identify inactive companies with growth potential and generate targeted plays to re-engage and revive pipeline," and then action those strategies in the Customer Success Workspace within HubSpot to drive retention. Support teams are offered the option to ask: "analyse seasonal patterns in ticket volume by category to forecast support team staffing needs for the upcoming quarter," and activate the Breeze Customer Agent in HubSpot to help manage anticipated support volume spikes. Nate Gonzalez, Head of Business Products at OpenAI, stated: "Launching the HubSpot deep research connector means businesses and their employees get faster, better insights because ChatGPT has more context. We're thrilled to work together to bring powerful AI to many of today's most important workflows." Colin Johnson, Senior Manager, CRM at Youth Enrichment Brands, commented: "The HubSpot connector is like having an extra analyst on the team, empowering sales reps to identify risks, opportunities, and next best actions. For a non-technical user, the fact that it's easy to use and talks directly to my data is huge." Karen Ng, SVP of Product and Partnerships at HubSpot, said: "We're building tools that help businesses lead through the AI shift, not just adapt to it. By connecting HubSpot CRM data directly to ChatGPT, even small teams without time or data resources can run deep analysis and take action on those insights - fueling better outcomes across marketing, sales, and service." The setup process is described as straightforward. HubSpot customers with administrative privileges can enable the connector by accessing ChatGPT, selecting the HubSpot deep research connector, choosing HubSpot as a data source, and authenticating their account. Post setup, any user within the organisation can toggle the feature on, sign in, and begin querying their data through natural language. HubSpot emphasised security considerations in the deployment of the connector. Users will only have access to CRM data permitted by their roles within HubSpot. Individual sales representatives, for example, will only be able to view pipeline data for deals where they have permission. The connector will be available automatically to all eligible HubSpot customers with paid ChatGPT plans, with coverage in the European Union including Team, Enterprise, and Edu subscriptions, and in other regions including Team, Enterprise, Pro, Plus, and Edu subscriptions. ChatGPT's language response is dictated by the language used in prompts.
Techday NZ
4 days ago
- Techday NZ
How Norby the language robot is more than just a companion
Norby, an AI-powered language companion, highlights the exciting evolution of language learning through interactive and adaptive technology. Adrian Mullan, Founder and CEO of Norby, led the demonstration at Dell Technologies World in Las Vegas, highlighting the robot's ability to engage in natural conversations across multiple languages and contexts. Norby supports more than 30 languages, with new additions and updates rolled out through over-the-air software updates. Mullan said his motivation for developing Norby was rooted in his own experiences. "A few years ago, while helping my daughter with her homework, I noticed something troubling - she was bored and struggling to stay focused. Conventional learning materials simply couldn't hold her attention. That realization sparked the idea for Norby," he explained. Norby is designed to be more than just another digital assistant or educational tool. "Humans have an innate desire to connect, even with the things we create. We give objects human-like traits, attribute emotions to our pets, and name our digital assistants - it's part of who we are," Mullan said. "I didn't want another device that demanded my attention. I wanted a companion - something that helps me stay focused, in flow, and truly engaged. A companion that adapts to my personality, interests, and goals." During the demonstration, Mullan illustrated Norby's capabilities by asking it to greet the audience in various languages and accents, as well as to generate stories on demand. "Hey, Norby, can you do me a favour? Can you say hello to the journalists in the room?" Mullan asked. Norby responded with, "Hello everyone. I'd be happy to extend a greeting to the journalists present. I'm sure, Adrian, you'd like me to make a good impression, so I'll make sure to be as charismatic as possible on various projects, including sound checks and helping with your work on interactive voice companions." The device's multilingual proficiency is underpinned by a flexible architecture that allows Norby to select the best available models for different languages. Mullan explained, "We plug into a bunch of different providers for the speech recognition as well as the large language model inference, and effectively what we do is we built an architecture where we can send the responses to the best model at the time." "So an example would be OpenAI, Google, Anthropic, all the big US providers will be great when it comes to English, Latin based languages and so on, but they'll tend to struggle with Asian character sets, so Chinese, Japanese, Korean and so on. In that case, we can spin it off to the DeepSeek, or to Qwen , or whoever, the best provider is," he said. The hardware is intentionally designed to be approachable and engaging, resembling a character rather than a device, and is equipped with high-performance processing, a full-colour HD touchscreen, high-quality speakers, and expressive RGB LED lighting. Norby's software supports a variety of large language models and is continually enhanced through regular updates. Norby's applications go beyond language learning. "Our problem at the moment is we have more requests from people than we can sort of deal with. I would say language and speech therapy is a really big one on the consumer side. The next area that we're moving into is much more the agentic productivity side," Mullan said. He described using Norby at home to manage daily tasks: "My Norby unit at home actually does things. I checks my email. It'll summarise my emails. I can ask Norby, okay, what's important for the day. And then what it'll do is, it's connected to my inbox, but also to my LinkedIn, to my social feeds, to my calendar. It'll take all of that stuff, create a summarisation." Addressing the broader technology landscape, Mullan compared Norby's form factor with other AI devices. "If you have something that's like projecting onto your hand, like, what happens when you walk out in daylight, right? So I think that's in their case, a big problem." "In the case of rabbit, I think they're doing some interesting stuff. Again, though my issue is, is that I think the moment you make a device that's very similar to a phoney form factor, you're effectively competing with the phone. And then in order to be able to move forward, what happens is, your expectation is that people have to carry this device and also carry the phone." "So we've made a deliberate decision not to do that. We've gone after a form factor that's a bit more anthropomorphised," he explained. Norby's adaptability is reflected in its ability to switch between personalities and modes depending on user needs. Parents, for example, can select predefined personas for their children, and the robot adapts its lessons to maintain engagement. "As the kids interact with Norby, it will course correct. So an example would be, okay, any kid that's learning a language, one of the challenges is engagement, right? Because they find it hard and boring. If it knows that, okay, there's an eight year old boy named Billy who's really into Roblox and Spider Man, it will then start to recreate those lessons in that context," Mullan said. Discussing privacy and security, Mullan stressed, "We take privacy and security seriously. Norby has been designed to be COPPA compliant. Our closed-system architecture means no access to third-party app stores or social media." Looking ahead, Mullan outlined the expansion of Norby's functionality, including the introduction of features such as singing, laughing, and emotional intelligence analysis. "There's a version that we're working on now that will sing, laugh, cry, whisper, yeah. I've had a chance to interact with that one. It's not production-ready yet, but it will happen," he said. "We're thrilled to be at this stage, and we're excited about the potential for Norby to make a difference in people's lives."
Techday NZ
5 days ago
- Techday NZ
How Norby the language robot is more than just a compaion
Norby, an AI-powered language companion, highlights the exciting evolution of language learning through interactive and adaptive technology. Adrian Mullan, Founder and CEO of Norby, led the demonstration at Dell Technologies World in Las Vegas, highlighting the robot's ability to engage in natural conversations across multiple languages and contexts. Norby supports more than 30 languages, with new additions and updates rolled out through over-the-air software updates. Mullan said his motivation for developing Norby was rooted in his own experiences. "A few years ago, while helping my daughter with her homework, I noticed something troubling - she was bored and struggling to stay focused. Conventional learning materials simply couldn't hold her attention. That realization sparked the idea for Norby," he explained. Norby is designed to be more than just another digital assistant or educational tool. "Humans have an innate desire to connect, even with the things we create. We give objects human-like traits, attribute emotions to our pets, and name our digital assistants - it's part of who we are," Mullan said. "I didn't want another device that demanded my attention. I wanted a companion - something that helps me stay focused, in flow, and truly engaged. A companion that adapts to my personality, interests, and goals." During the demonstration, Mullan illustrated Norby's capabilities by asking it to greet the audience in various languages and accents, as well as to generate stories on demand. "Hey, Norby, can you do me a favour? Can you say hello to the journalists in the room?" Mullan asked. Norby responded with, "Hello everyone. I'd be happy to extend a greeting to the journalists present. I'm sure, Adrian, you'd like me to make a good impression, so I'll make sure to be as charismatic as possible on various projects, including sound checks and helping with your work on interactive voice companions." The device's multilingual proficiency is underpinned by a flexible architecture that allows Norby to select the best available models for different languages. Mullan explained, "We plug into a bunch of different providers for the speech recognition as well as the large language model inference, and effectively what we do is we built an architecture where we can send the responses to the best model at the time." "So an example would be OpenAI, Google, Anthropic, all the big US providers will be great when it comes to English, Latin based languages and so on, but they'll tend to struggle with Asian character sets, so Chinese, Japanese, Korean and so on. In that case, we can spin it off to the DeepSeek, or to Qwen , or whoever, the best provider is," he said. The hardware is intentionally designed to be approachable and engaging, resembling a character rather than a device, and is equipped with high-performance processing, a full-colour HD touchscreen, high-quality speakers, and expressive RGB LED lighting. Norby's software supports a variety of large language models and is continually enhanced through regular updates. Norby's applications go beyond language learning. "Our problem at the moment is we have more requests from people than we can sort of deal with. I would say language and speech therapy is a really big one on the consumer side. The next area that we're moving into is much more the agentic productivity side," Mullan said. He described using Norby at home to manage daily tasks: "My Norby unit at home actually does things. I checks my email. It'll summarise my emails. I can ask Norby, okay, what's important for the day. And then what it'll do is, it's connected to my inbox, but also to my LinkedIn, to my social feeds, to my calendar. It'll take all of that stuff, create a summarisation." Addressing the broader technology landscape, Mullan compared Norby's form factor with other AI devices. "If you have something that's like projecting onto your hand, like, what happens when you walk out in daylight, right? So I think that's in their case, a big problem." "In the case of rabbit, I think they're doing some interesting stuff. Again, though my issue is, is that I think the moment you make a device that's very similar to a phoney form factor, you're effectively competing with the phone. And then in order to be able to move forward, what happens is, your expectation is that people have to carry this device and also carry the phone." "So we've made a deliberate decision not to do that. We've gone after a form factor that's a bit more anthropomorphised," he explained. Norby's adaptability is reflected in its ability to switch between personalities and modes depending on user needs. Parents, for example, can select predefined personas for their children, and the robot adapts its lessons to maintain engagement. "As the kids interact with Norby, it will course correct. So an example would be, okay, any kid that's learning a language, one of the challenges is engagement, right? Because they find it hard and boring. If it knows that, okay, there's an eight year old boy named Billy who's really into Roblox and Spider Man, it will then start to recreate those lessons in that context," Mullan said. Discussing privacy and security, Mullan stressed, "We take privacy and security seriously. Norby has been designed to be COPPA compliant. Our closed-system architecture means no access to third-party app stores or social media." Looking ahead, Mullan outlined the expansion of Norby's functionality, including the introduction of features such as singing, laughing, and emotional intelligence analysis. "There's a version that we're working on now that will sing, laugh, cry, whisper, yeah. I've had a chance to interact with that one. It's not production-ready yet, but it will happen," he said. "We're thrilled to be at this stage, and we're excited about the potential for Norby to make a difference in people's lives."
