AI News · The messy stuff ·
Protegrity study says 82% of leaders push AI agents out in a diminished state
Research commissioned by data security firm Protegrity from Enterprise Management Associates found that 83% of leaders reported AI project delays, SiliconANGLE reported. According to Protegrity's Diego Loureda, the same study found that 82% push agents into production stripped of the tools and data that would make them useful. Loureda argued that approving AI projects one by one will not scale.
Key points
- A Protegrity commissioned EMA study found 83% of leaders reported AI project delays.
- Loureda said 82% push agents to production without the tools and data they need.
- Protegrity argues companies should approve reusable patterns rather than individual AI projects.
What happened: Enterprises moving AI agents into production are finding that data protection now decides how fast those projects move, SiliconANGLE reported from Cloudera's EVOLVE26 event. The report drew on an interview with Diego Loureda, vice president of customer success and presales at Protegrity, which Cloudera named its Emerging Technology Partner of the Year.
The details: According to research Protegrity commissioned from Enterprise Management Associates, 83% of leaders reported project delays. Loureda said the same study found that 82% of leaders push agents into production in a diminished state, stripped of the tools and data that would make them useful. The sample size, date and method of the study were not reported. Loureda warned that "data leaks are still a thing" when sensitive data ends up in a model, a vector store or an agent's memory.
Background: SiliconANGLE said traditional encryption can lock data so tightly that analysts and models can no longer use it. Protegrity focuses on tokenization, which swaps sensitive values for substitutes. Loureda said the company's tokenization is "already quantum ready."
What to watch: Loureda said the main governance problem is that companies approve individual projects, which "will not scale in the age of AI," and that they "have to approve patterns" instead. He added that security is now "the team that can enable AI to give you value." SiliconANGLE disclosed that theCUBE was a paid media partner for the event.
Our take
Teams rolling out AI agents should settle data protection and approval patterns early, or risk shipping agents too limited to deliver value.