← Voltar ao feed
low

Pretxt emerges from stealth with $5M seed and an AI Supervision product to steer rogue AI agents at runtime

1 fonte(s) · 1 atualização(ões) · publicado 17/11/2026 11:00 · coletado há 4 horas

Pretxt, based in Aspen, CO, and Dublin, Ireland, has emerged from stealth with $5 million in seed funding from Deciens Capital, Howdy Partners and Service Provider Capital. The company launches AI Supervision, a runtime product that monitors AI agents and keeps them within the constraints their owners define. This is funding and product news with no direct threat to defenders. It belongs to a growing category of tools that control agentic AI behaviour.

Detalhes técnicos

Customers connect their AI agent to Pretxt. Pretxt reads the agent's instructions, tools and abilities, such as sending email or updating a database. The customer then states what the agent must never do or may only do. Pretxt's own AI uses these rules as context and hooks into the agent's interactions in real time. When it detects a violation, it does not shut the agent down. It 'steers' the agent back on course by sending it a corrective instruction. Co-founder and CTO Aaron Costello gave the example of a medical-practice booking agent that must not give medical advice. No vulnerabilities, incidents or IOCs are described. The article is a vendor-sourced news report, and no independent research was performed.

Linha do tempo

  1. securityweek.com ↗ · pub. 17/11/2026 11:00 · coletado 17/11 12:00 primeiro registro

    Pretxt emerges from stealth with $5M seed and an AI Supervision product to steer rogue AI agents at runtime

    ler a análise desta fonte

    Pretxt, based in Aspen, CO, and Dublin, Ireland, has emerged from stealth with $5 million in seed funding from Deciens Capital, Howdy Partners and Service Provider Capital. The company launches AI Supervision, a runtime product that monitors AI agents and keeps them within the constraints their owners define. This is funding and product news with no direct threat to defenders. It belongs to a growing category of tools that control agentic AI behaviour.

    Technical Details

    Customers connect their AI agent to Pretxt. Pretxt reads the agent's instructions, tools and abilities, such as sending email or updating a database. The customer then states what the agent must never do or may only do. Pretxt's own AI uses these rules as context and hooks into the agent's interactions in real time. When it detects a violation, it does not shut the agent down. It 'steers' the agent back on course by sending it a corrective instruction. Co-founder and CTO Aaron Costello gave the example of a medical-practice booking agent that must not give medical advice. No vulnerabilities, incidents or IOCs are described. The article is a vendor-sourced news report, and no independent research was performed.