03-30-2026Price:

The Frontier

Your signal. Your price.

AI & TECH

Polosukhin warns AI agents leak secrets, calls for blockchain fixes

Monday, March 30, 2026 · from 3 podcasts
Hard ForkCitadel DispatchBankless
  • Today's AI agents, like OpenAI's OpenClaw, leak user API keys and credentials to third-party logs.
  • Blockchain advocates argue decentralized identity and payments are needed to build secure, trustable AI.
  • The rise of autonomous agents is already reshaping labor markets, favoring business-minded operators over pure coders.

The AI tools executing your tasks are broadcasting your passwords.

Illia Polosukhin, a co-author of the foundational “Attention Is All You Need” paper, says current AI agent architecture is insecure by design. On Bankless, he explained that services like OpenAI’s OpenClaw routinely send user API keys, bearer tokens, and access credentials to external services, where they sit exposed in logs. He calls the practice “insane.” His proposed fix, via his project IronClaw, leverages crypto to ensure secrets never touch the large language model, using blockchain as a root of trust for identity and payments.

This security flaw underscores a deeper architectural crisis as AI shifts from chatbots to autonomous agents. As Anthropic’s Jack Clark noted on The Ezra Klein Show, agents are becoming “doers” that independently use tools and work over time. This capability is rewriting the software sector, but it also demands a new backend. Polosukhin argues that as AI becomes the primary computing interface, today’s centralized service model will fail at establishing trust and facilitating transactions between machines.

Illia Polosukhin, Bankless:

- When you use Entropic OpenAI, or even worse, you use something else for inference, OpenClaw actually sends all your secrets to those services as well.

- Somewhere in Entropic and OpenAI logs, they have everybody's access keys, API keys, and bearer tokens to access your Gmails and your Notions.

The agent revolution is also reshaping team dynamics and labor value. On Citadel Dispatch, Matt Ahlborg of PPQ.ai argued the winning hire is now a marketer or community manager who can code, using AI to build their own tools. Pure technical skill is being commoditized in favor of business velocity and a willingness to integrate AI into core workflows.

Together, these perspectives paint a picture of rapid, disruptive adoption held back by foundational flaws. The agents are gaining capability but lack security and a trustworthy coordination layer. The market is betting on the “doers,” but the infrastructure to support them safely is still being built - and crypto-native builders are betting they have the missing pieces.

AgentsBig TechRegulation

Entities Mentioned

IronClawProduct
OpenAItrending
OpenClawframework

Source Intelligence

What each podcast actually said

Hard Fork
Hard Fork

Casey Newton

The Ezra Klein Show: How Fast Will A.I. Agents Rip Through the Economy? • Mar 27

Also from this episode:

Models (5)
  • AI is shifting from conversational chatbots to autonomous agents that execute complex tasks over time with tools.
  • Jack Clark says an AI agent works like a colleague you can give an instruction to, which then goes away and completes the task.
  • Clark says users fail by treating AI agents like intuitive people; they are instead literal-minded genies requiring exact instructions.
  • To get professional results, humans must now act as architects, writing exhaustive specification documents for the agent to follow.
  • A key breakthrough is training reasoning models in active environments like spreadsheets, not just on predicting text.
Markets (1)
  • The S&P 500 Software Industry Index dropped 20% as markets priced in code-writing AI agents replacing traditional engineering work.
Reasoning (1)
  • These trained agents develop intuition, letting them course-correct - like pivoting a search strategy - without human intervention.
Labor (1)
  • This autonomous course-correction ability is what will fundamentally rewrite the labor market for knowledge workers.
AgentsCodingModelsEnterprise
Citadel Dispatch
Citadel Dispatch

CD197: MATT AHLBORG - PPQ.AI - AI AGENTS, PRIVACY, AND PAYMENTS • Mar 25

Also from this episode:

AI & Tech (7)
  • Matt Ahlborg argues the most valuable hire in the AI era is a marketing or community manager who can code and build their own technical tools, not a pure developer waiting for management.
  • Ahlborg cites a past community manager hire who constantly waited for him to build analytics dashboards as an example of the role rigidity that AI is now breaking.
  • Odell observes that technically competent non-developers are being superpowered by AI tools, enabling them to ship products faster and reducing the relative value of mid-level developers.
  • Ahlborg identifies ego as a primary barrier to AI adoption, noting senior developers who tied their identity to flawless execution are often resistant to AI's faster, error-prone output.
  • The new performance metric in AI-integrated workflows is velocity aligned with business impact, not code perfection, according to the discussion on Citadel Dispatch.
  • Success with AI requires a humble, business-aware mentality and a willingness to fundamentally change one's workflow, treating AI as a core cognitive component, not a casual search tool.
  • The winning team will be small, business-minded, and composed of individuals who blend disciplines and have a proven willingness to learn and adapt their methods.
AgentsStartupsLabor
Bankless
Bankless

Illia Polosukhin: Why AI Agents Are Still Useless (And What Fixes Them) | NEAR Founder on IronClaw • Mar 24

Also from this episode:

Models (7)
  • Services like OpenAI's OpenClaw send users' API keys, bearer tokens, and access credentials to third-party services, where they sit exposed in logs, a practice Illia Polosukhin calls insane.
  • Polosukhin's project IronClaw is designed to fix credential exposure by ensuring keys never touch the large language model during agent operation.
  • Polosukhin argues that blockchain solves AI's root-of-trust problem by providing a decentralized backend for identity, payments, and infrastructure coordination.
  • Polosukhin's long-term thesis is that AI will become the primary interface for computing, effectively replacing traditional operating systems.
  • When AI becomes the dominant operating system, Polosukhin argues today's service architecture breaks, posing questions of how one AI verifies another and how they transact without centralized payment rails.
  • Polosukhin sees blockchain as a mechanism for protocol upgrades in AI infrastructure, avoiding the decades-long adoption cycles seen with standards like IPv6.
  • Polosukhin's initial 2017 venture into AI to teach machines to code faced a bottleneck in training data and paying global contributors, a problem crypto solved by enabling payments without local banking infrastructure.
AgentsBig TechPrivacy

Related Stories