Follow ZDNET: Add us as a preferred source<!–> on Google.
ZDNET’s key takeaways
- ServiceNow and Nvidia have launched a new open-source AI model.
- It’s designed to build custom agents and optimize security.
- Despite ample hype, many businesses aren’t seeing ROI from AI.
Software company ServiceNow announced on Tuesday that it’s expanding its partnership with US chipmaker Nvidia with the launch of a new open-source AI model named Apriel 2.0, designed specifically to help businesses build custom agents.
Also: Claude can integrate with Excel now – and gets 7 new connectors
If you’re new here, agents are AI systems that can access an array of digital tools and perform multistep tasks with little to no direct human oversight. In a press release, ServiceNow said that Apriel 2.0 blends “enhanced reasoning” with multimodal input capabilities (meaning it can process text, audio, images, and other data formats), and that it “matches the reasoning and accuracy of much larger models at a fraction of the size.”
What it offers
Apriel 2.0 is the successor to the Apriel Nemotron 15B, another open-source, enterprise-facing model that debuted earlier this year. It’s built upon Nvidia’s Nemotron open model architecture, meaning developers can access its underlying code to build their own agents on top of it. It also operates within the ServiceNow AI Platform, allowing current ServiceNow customers to transfer their current data governance policies into the new model.
–>
“The model will run within ServiceNow’s secure, compliance-certified infrastructure,” Joe Davis, Executive Vice President, Platform Engineering and AI, told ZDNET. “Because it will be embedded in [Now LLM], Apriel 2.0 will automatically inherit each organization’s existing permissions and audit trails, ensuring AI outputs stay within the same policies that already govern their business.
Like the Apriel Nemotron 15B, Apriel 2.0 — which is expected to go into production starting in the first quarter of next year — will be available via Hugging Face.
Security as a selling point
ServiceNow and Nvidia are selling their new model to businesses as a safer and more secure alternative to the general-purpose agentic AI tools that are currently on the market.
Also: 50 AI agents get their first annual performance review – 6 lessons learned
“Open models give enterprises the transparency and control they need to specialize AI to their data, workflows, and trust standards,” Kari Briski, Vice President of Generative AI for Enterprise at Nvidia, said in a statement.
The companies are also specifically hoping to sell the new model to federal agencies and businesses working in tightly regulated industries like healthcare, finance, and telecommunications. The industry is responding to security concerns elsewhere as well; for example, Snowflake recently added MCP support to its AI solutions for finance.
Many businesses seem to be warming up to the idea of using agents, but mostly this has been limited to narrow, low-stakes tasks.
Also: Why open source may not survive the rise of generative AI
Agents also present their own data privacy and security risks: greater agency, for example, usually requires users to hand over greater quantities of their proprietary data, leading to the danger of that data being accidentally leaked or even inexplicably deleted.
According to ServiceNow’s press release, Apriel 2.0 combines “efficient performance with responsible data vetting, safety guardrails and transparency controls,” making it a safer option for federal agencies and businesses working across regulated sectors.
Where’s the ROI?
The news arrives at a time when many businesses are struggling to achieve measurable ROI with their internal AI initiatives, and when consumer AI chatbots seem to be presenting, in some cases, more cybersecurity risks than tangible increases to employee productivity.
Also: 3 tips for navigating the open-source AI swarm – 4M models and counting
Meanwhile, agents have become a major focus for tech developers looking to turn a profit on AI. The market research firm Gartner recently forecasted that as much as half of all internal decision-making processes within businesses could be (at least partially) automated by agents within the next two years.
Want more stories about AI? Sign up for AI Leaderboard, our weekly newsletter.
Artificial Intelligence
<!–>
–>
