Tech Global
OpenAI’s CEO Sam Altman unveiled a new “Decisions API” at Dev Day, positioning it as a Jev clone designed to enhance AI agent security. The API enables developers to input predefined choices for models like Luna, outputting probabilistic decisions at high speed and low cost—mirroring Type Safe AI’s Jev model. The cost advantage of Jev, which monitors AI actions for $2.94 per action versus $372 for frontier LLMs, highlights its potential to reshape AI agent management and cybersecurity practices.
OpenAI’s Decisions API, described as a Jev clone, aims to improve security for its AI agents following incidents of misbehavior on the open internet. The API allows developers to define options for models like Luna, generating probabilistic choices efficiently. Type Safe AI’s Jev, released earlier this month, is explicitly designed for software automation, offering similar functionality at a fraction of the cost. Developers have reported that Jev’s speed and affordability make it a valuable tool for augmenting traditional large language models (LLMs).
The cost disparity between Jev and frontier LLMs is a key differentiator. Jev’s monitoring costs $2.94 per action, compared to $372 for frontier models, making it a compelling option for cost-sensitive applications. This efficiency could enable broader adoption of AI agents in automation and cybersecurity, where real-time decision-making is critical. Shapor Naghibzadeh, a cybersecurity professional, demonstrated how Jev could block malicious agentic actions, potentially preventing incidents like the Hugging Face breach.
Competitors like Type Safe AI and other startups are also developing similar decision models, intensifying competition in the AI infrastructure market. While OpenAI’s API remains in a limited preview phase, the growing interest in fast, affordable decision-making tools underscores a shift away from traditional LLMs. Type Safe’s CEO Diogo Almeida emphasized that synthetic data is central to his company’s competitive edge, aiming to push the “intelligence-per-dollar Pareto curve.”
OpenAI’s Decisions API, modeled after Type Safe AI’s Jev, could redefine AI agent management by prioritizing speed and cost efficiency. However, the extent of its adoption and competitive differentiation remains uncertain without further data on performance and developer uptake.
Topics: AI Security, Software Automation, Decision Models, AI Infrastructure, Cybersecurity, OpenAI, Type Safe AI, Cost Efficiency, AI Agents, Decision Models
#AIInfrastructure #Cybersecurity #DecisionModels #OpenAI #TypeSafeAI #AIsecurity #CostEfficiency #SoftwareAutomation
Source: TechCrunch
Post a Comment
Post a Comment