At eMerge Americas, AE Studio, Zylon AI and GovSignals described AI risk efforts underway before Dario Amodei warned AI could outpace human control: alignment research, private deployments for regulated clients and software that can switch among language models.
McKinsey’s 2026 AI Trust Maturity Survey found nearly two-thirds of respondents cited security and AI risk concerns as the leading obstacle to scaling agentic AI. Deloitte reported that 74% of business and IT leaders expected at least moderate use of AI agents by 2027, while only 21% of organizations said their governance of autonomous agents was mature.
AE Studio’s DARPA-funded AICRAFT program pairs outside researchers in alignment, control and interpretability with engineering teams for two-week sprints to explore safety beyond surface-level filters. Zylon runs open-weight models on customer-controlled servers or private clouds, though hardware, implementation and maintenance still carry costs. GovSignals says customer data stays within its FedRAMP High-authorized, Defense Department Impact Level 5 environment and is not used to train external providers’ shared models.
