Welcome to AI Confidential, your biweekly breakdown of the most interesting developments in confidential AI.

Today we’re exploring:

• Enterprise-level use cases for GenAI

• ChatGPT’s sycophant-y mood update

• Open source projects we’ve been following

Also mentioned in this issue: James Kaplan, Derek Manky, John Willis, João Moura, Atindriyo Sanyal, Jake Broekhuizen, Jason Clinton, Nelly Porter, Mark Russinovich, David Reber, Mark Papermaster, McKinsey & Company, Gartner, CB Insights, IDC Marketscape, crewAI, Galileo, LangChain, BCG, ChatGPT, Deloitte, Cyberhaven, Netwirx, Cisco, Kimi-Audio, Moondream, and FortiGuard Labs.

Let’s dive in!

This week on the AI Confidential podcast, we continue with a decisive look at how companies can actually capture value with generative AI.

In this episode, OPAQUE CEO Aaron Fulkerson sits down with James Kaplan, Partner and CTO of McKinsey Technology, to discuss the power of GenAI adoption across multiple different industries.

Having spent the past 25 years helping the world’s largest companies integrate cutting-edge technology into the core of their businesses, James’ position on GenAI is clear:

It isn’t just a tool for individual users.

It has huge potential to help IT teams modernize faster, innovate further, and be more effective than ever.

GenAI is no longer just for writing blog outlines and conducting simple research. It’s becoming a key player in high-value enterprise-level opportunities like:

• Scanning messy code to find patterns and insights

• Migrating outdated, legacy systems into the cloud

• Creating modernized apps with increased security and resiliency 

Like with any other new tech, James knows that companies that capitalize on GenAI are going to reap the benefits.

They just need to foster a healthy innovation metabolism across both platform and use cases to get the most out of their efforts.

In this episode, you’ll also hear James share insights into:

• Why IT is cheap—but making it work at scale isn’t

• How to break free from the “IT doom loop” (yes, it’s real)

• Why a clear AI strategy matters more than ever

The episode is live now. You can subscribe and listen here.

Keeping It Confidential

By 2027, what percentage of data breaches are expected to be caused by improper use of GenAI tools across international borders?

1. Under 5%

2. 10%

3. 25%

4. Over 40%

Answer at the bottom of this newsletter.

Code for Thought

Important AI news in <2 minutes

⚠️ 83.8% of company data is being fed into risky AI tools that could leak, a report from Cyberhaven finds.

🔓 Generative AI tools are becoming increasingly vulnerable to jailbreaks, injection attacks, and data theft as cyber threats get more sophisticated, reports Hacker News.

💸 75% of enterprises faced financial damages because of cyber attacks—up 15% since 2024, according to Netwrix.

📉 66% of board members say they have little to no knowledge or experience using AI, a new Deloitte report finds.

🧠 ChatGPT’s latest model seems to have a personality problem (TL;DR: it became a sycophant).

🏢 Employees at big consulting firms like McKinsey and BCG are going all-in on generative AI to do their jobs better and faster.

Community Roundup

Updates involving OPAQUE and our partners

🚀 The countdown is on — OPAQUE’s 2025 Confidential Computing Summit is just over a month away!

Join us from June 17 to 18 in San Francisco for two action-packed days of breakthrough talks, lively conversations, and unbeatable networking. 

🎤 Featured speakers include:

• Jason Clinton, CISO, Anthropic

• Nelly Porter, Director of Product Management, Google

• Mark Russinovich, CTO, Microsoft Azure

• David Reber, CSO, NVIDIA

• Mark Papermaster, CTO, AMD

Still want more? Join our pre-Summit workshop, The New NORMAL: Normalizing AI Enterprise Architecture for AI, on June 16th to hear from more insightful speakers, including:

• John Willis, author of Rebels of Reason and The DevOps Handbook

• João Moura, CEO, crewAI

• Atindriyo Sanyal, CTO, Galileo

• Jake Broekhuizen, Deployed Engineer, LangChain

🔒 This is where the leaders of confidential computing come to connect—and where we explore the next frontier of AI.

Tickets are going fast. Secure yours now — before it’s too late.

OPAQUE in the press

Exciting news: CB Insights has chosen OPAQUE as one of the world’s top 100 promising private AI companies!

This recognition confirms everything our team knows to be true:

• Whoever controls the data controls the industry

• There is a demand for confidential AI solutions

• AI innovators need to leverage sensitive data

• Our work is truly just getting started

Making a market splash

The IDC MarketScape: Worldwide Data Clean Room Technology 2025 Vendor Assessment is out now, and OPAQUE is proud to be named a major player.

As one of the most well-trusted sources of market insights, IDC listed the following areas as our core strengths:

⭐️ Confidential computing technology

⭐️ Verifiable audit trail

⭐️ Enhanced identity resolution

⭐️ Flexible deployment

Thanks to the IDC team for including us in this report!

OPAQUE in the wild

Last week, OPAQUE’s Vice President and Head of Engineering, Giorgio Natili, was a guest on the Healthy Mind podcast 🎙️

While the episode focuses on leadership, it also touches on:

• How AI should augment humans, not replace them

• Why data privacy is the foundation of responsible innovation

• The ethical “sharks” leaders need to stare down

If you care about AI, ethics, and real-world leadership, you’ll want to give it a listen.

Open source spotlight 

🎵 Kimi-Audio is a foundational LLM specializing in audio processing, like speech recognition, audio question answering, closed captioning, and speech emotion recognition. 

🔐 Cisco released Foundation-sec-8B, a foundational AI model designed to address security by fine-tuning cyber risk protocols and detecting organizational threats.

👁️ Moondream is a visual language LLM that comprehends images using simple text prompts, including captioning, visual question answering, and object detection.

Quotable

🛡️“The traditional security playbook is no longer enough. Organizations must shift to a proactive, intelligence-led defense strategy powered by AI, zero trust, and continuous threat exposure management to stay ahead of today’s rapidly evolving threat landscape.” 

— Derek Manky, Chief Security Strategist and Global Vice President for Threat Intelligence at Fortinet’s FortiGuard Labs (Source: Fortinet)

Trivia answer: Over 40%.

The adoption of GenAI tools is outpacing data governance protocols worldwide. As a result, Gartner predicts that more than 40% of AI-related data breaches will be caused by improper GenAI practices by 2027. Yikes!

Stay confidential,

- Your friends at OPAQUE

ICYMI: Links to past issues

• The Untapped Axes of AI Innovation

• Unleashing the Power of Agents

• Why Data Masking Isn’t Working for Your AI