Should your security budget double?

Your security budget may need to double

a16z's infrastructure team sent its portfolio companies a short, blunt memo after last week’s wave of AI-found vulnerabilities. Security flaws are now being found and exploited faster than ever before. On top of that, every AI agent we put to work is a potential insider threat, which is something our defences were never designed for.

While their advice is a bit alarmist, it's hard to argue against. We do need to increase the spend and focus on security within our infrastructure. Their recommendations are to consider doubling security from the typical 10% of tech spend, delay new launches until known holes are patched, switch off the neglected test and legacy systems that earn no revenue, and stop treating risk acceptance as a polite way of never fixing anything. They also say to grow the security team with agents rather than headcount, because attackers are already using agents to find their way into systems.

In my experience, the launch delay is the hardest one to sell, because it lands on the product roadmap and the revenue plan at the same time. But I’d much rather have that argument now than after a breach. Setting up teams empowered with agents to run penetration tests and patch holes across our organisations is the best way to get ahead of attackers.

Also worth your time

Model Behavior: Rick Rubin, Substack. Substack founder Chris Best asks producer Rick Rubin where he lands on AI in creative work. Rubin treats it as a new instrument, and tells the story of a guitarist who binned one of the first wah-wah pedals while Jimi Hendrix built a whole sound on it. For those of us leading creative teams, it's a good case for leaving room to experiment before we draw any hard lines.

How to Get AI Agents to Recommend Your Articles, Every. Mike Taylor ran 2,484 test sessions to see which articles Claude and ChatGPT choose to cite. Titles that say plainly what is inside, and descriptions that read as an answer, won by a wide margin, and adding "2026" to a title did not help at all. A great one to send your CMO, now that AI assistants are becoming a front door to your brand.

Here is what women want from AI, Elena Verna. The New York Times recently ran a feature that framed what women want from AI around outfits and party planning. Verna tears this to shreds, arguing that AI can be used for so much more, from breaking down the glass ceiling to giving women more influence and access, and that, yes, taking over household responsibilities would help too. For our CHROs, it's a prompt to look at how we might reframe AI training and success stories to be less about saving time and more about empowering us to shoot for the moon.

2026 Board & CEO Insights Report, Eton Bridge Partners. The search firm surveyed 126 Chairs, CEOs and non-executives between July and September 2026, and just over half named governing AI as a board exposure, yet only 8% think their board adds real value on it. CEOs mostly worry about AI's cost, returns and losing ground to competitors, while non-executives worry about cyber and data risk, so it's well worth getting both views into the same boardroom conversation.

Is your leadership team ready to light the fire?

Upskill with AI Firestarter bootcamps and workshops, built for leaders who are too busy leading to be experimenting with AI.

As a leader, you set the tone for how AI is used in your organisation.

Until next time,
Ryan Snyder
ryan@aifirestarter.com