New Partnership Verona x Atbash
AI agents are now capable of understanding instructions, devising plans, writing code, and executing transactions. The main issue is no longer simply intelligence, but control: Is the information being used accurate, and are the intended actions actually permitted?
The partnership between $VERONA and ATBASH addresses these two questions from different angles. Verona ensures that the information used by the agent has been verified directly from the source. This information is then converted into a proof that can be reused by authorized agents, without them having to access the user’s raw data.
However, operating based on accurate information does not necessarily result in permitted actions. An agent might have accurate data but still attempt to transfer funds beyond the limit, access sensitive systems, or execute code that should require approval. This is where Atbash comes into play.
Atbash sits between the agent and the tools it can access. Companies can set their own "red lines" or operational boundaries. Before an irreversible action is executed, Atbash verifies the agent’s identity, the target it seeks to access, applicable policies, and the possibility of prompt injection. The result can be authorization, blocking, or escalation for further review.
Atbash does more than just stop rule-breaking actions. Every decision also leaves a record of what the agent did, which rules were applied, and the reasons why the action was permitted or blocked. When security teams, customers, or regulators request an explanation later on, the decision log is readily available for review.
This collaboration expands Verona’s role in building AI agents suitable for real-world use. Verona secures information before decisions are made, while Atbash controls the actions resulting from those decisions. The result is not just a smarter agent, but one that operates based on facts, understands the limits of its authority, and can be held accountable.
AI agents are now capable of understanding instructions, devising plans, writing code, and executing transactions. The main issue is no longer simply intelligence, but control: Is the information being used accurate, and are the intended actions actually permitted?
The partnership between $VERONA and ATBASH addresses these two questions from different angles. Verona ensures that the information used by the agent has been verified directly from the source. This information is then converted into a proof that can be reused by authorized agents, without them having to access the user’s raw data.
However, operating based on accurate information does not necessarily result in permitted actions. An agent might have accurate data but still attempt to transfer funds beyond the limit, access sensitive systems, or execute code that should require approval. This is where Atbash comes into play.
Atbash sits between the agent and the tools it can access. Companies can set their own "red lines" or operational boundaries. Before an irreversible action is executed, Atbash verifies the agent’s identity, the target it seeks to access, applicable policies, and the possibility of prompt injection. The result can be authorization, blocking, or escalation for further review.
Atbash does more than just stop rule-breaking actions. Every decision also leaves a record of what the agent did, which rules were applied, and the reasons why the action was permitted or blocked. When security teams, customers, or regulators request an explanation later on, the decision log is readily available for review.
This collaboration expands Verona’s role in building AI agents suitable for real-world use. Verona secures information before decisions are made, while Atbash controls the actions resulting from those decisions. The result is not just a smarter agent, but one that operates based on facts, understands the limits of its authority, and can be held accountable.
