Ocean Security CTO Oran Moyal: AI-Native Is Organizational, Not Technical
Oran Moyal, who co-founded Ocean Security in 2024, says AI-native firms change how every team works — and that enforced permissions, not instructions, are the real security controls.
By Amara Osei
4 min read
What's News
- Oran Moyal and Shay Shwartz co-founded Ocean Security in 2024.
- Moyal says a $20 generative-AI subscription now gives attackers nation-state phishing capabilities.
- Ocean Security's agentic platform investigates phishing, business email compromise and financial fraud.
- Moyal's background includes the Israel Defense Forces and VisibleRisk / Microsoft Security.
- Moyal predicts agentic operating models will dominate tech companies within three to five years.
A $20 generative-AI subscription now gives every attacker nation-state phishing capabilities, according to Oran Moyal, Co-Founder and CTO of Ocean Security. "Overnight, a $20 subscription gave every attacker nation-state capabilities," Moyal said, describing why he and Shay Shwartz launched the company in 2024.
Ocean Security builds an agentic email security platform that investigates targeted phishing, business email compromise and financial fraud. Its AI agents analyze communication context and sender intent to catch threats hiding inside legitimate business threads, trusted accounts and real services. Moyal's background spans cybersecurity R&D, offensive social engineering, and enterprise security, including roles in the Israel Defense Forces and at VisibleRisk / Microsoft Security.
What changed in the threat landscape?
A targeted spear phishing campaign used to require a dedicated team and weeks of manual research and preparation. Generative AI collapsed that cost curve, Moyal said, putting every enterprise employee within reach of devastating attacks. Legacy tools fail because they hunt for rule violations and anomalies — but these attacks live inside normal business communication where nothing looks abnormal.
Ocean's answer is AI-driven email security that analyzes context and sender intent in real time, much like a human investigator would, before a compromise occurs.
What does 'AI-native' actually mean?
Buying AI tools is technology adoption, not transformation, Moyal argued. "Real AI-native transformation is organizational, not technical," he said. The distinction shows in how leaders react to a bottleneck: the traditional reflex is hiring; an AI-native organization first asks how AI can speed up the entire function.
If you accelerate one department, the bottleneck just moves to the next one, and the company ends up moving at the pace of its slowest function. AI-native organizations change together, so information flows faster, feedback loops shorten and teams operate as one connected system. As AI takes on execution, the scarce human skills become setting direction, defining intent and applying judgment.
Where do AI productivity gains disappear?
Speeding up individual tasks doesn't speed up the organization if processes, dependencies and communication between teams stay the same. The most common bottlenecks aren't technical: unclear requirements cause rework, slow handoffs stall work between teams, and approvals pile up. "AI makes each step faster, but the time is lost between the steps," Moyal said.
His prescription: measure where time actually goes, redesign workflows end-to-end rather than task-by-task, and connect every AI effort to KPIs teams already own. Automation should become the default mindset.
How should agents validate their own work?
Moyal defined three layers for delegating work to agents: clear tasks, goals and context resolved upfront; self-validation against success criteria defined in advance; and humans as the final layer applying judgment, not fixing basic mistakes. "If an agent decides what 'done' means, it will grade its own homework," he said.
Effective guardrails include deterministic tests that verify specific parameters and policies, an independent reviewer agent with fresh context, and evidence — test results, screenshots, recordings — rather than a bare "pass" verdict.
What about security and accountability?
Instructions alone don't count as security controls, Moyal warned. Real guardrails must be enforced: strict access limits, verification that agents can't reach anything outside their scope, and continuous monitoring that blocks unplanned actions. Accountability starts with the operator, who defines what an agent can do alone versus what needs approval. A threat-response agent might quarantine suspicious messages autonomously but require review before blocking a sender organization-wide, expanding its boundaries as its track record grows.
This matters more because attackers are increasingly targeting AI agents, while security teams still lack a clear playbook for protecting them, Moyal said.
How do you measure real value?
Time saved is only the first signal. Moyal pairs it with quality metrics — rework, defects, post-delivery corrections — plus business expansion measured as new initiatives taken on without adding headcount, and end-to-end speed from prioritization decision to live product. In security, end-to-end speed is existential: attackers already use AI agents, so detection and response times must keep shrinking.
Over the next three to five years, Moyal predicts successful technology companies will move beyond basic automation toward agentic operating models, with leaner human teams directing outcomes and competing against companies many times their size.
Source: unite.ai
More from Amara Osei
Show full bio
Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.
2,184 articles