> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Databricks CEO: AI Safety Theater Is Making AI Less Safe
- URL: https://wire.fourthweb.ai/databricks-ceo-ai-safety-theater-is-making-ai-less-safe/
- Published: 2026-09-18T20:01:49.000Z
- Updated: 2026-09-18T20:01:51.000Z
- Description: The loudest voices in AI safety are accidentally making the field less safe. Databricks CEO Ali Ghodsi says existential AI risk is "close to zero" and the real threat is cybersecurity, where AI is dramatically shortening exploit timelines
- Author: Travis Wright
- Tags: AI Agent Economy, AI Governance, DeFi, Anthropic

**The loudest voices in AI safety are accidentally making the field less safe.**

### The Summary

- [Databricks CEO Ali Ghodsi says existential AI risk is "close to zero"](https://www.bloomberg.com/news/videos/2026-09-17/ai-existential-risk-is-close-to-zero-databricks-ceo-video?ref=wire.fourthweb.ai) and the real threat is cybersecurity, where AI is dramatically shortening exploit timelines
- [Critics argue existential risk warnings from companies like Anthropic are hijacking the AI debate](https://www.bloomberg.com/news/videos/2026-09-18/anthropic-s-existential-risk-warnings-hijack-ai-debate-video?ref=wire.fourthweb.ai), pulling attention from problems that already exist
- As AI starts building the next generation of AI, the gap between theoretical doom scenarios and present-day vulnerabilities is defining two camps in tech

### The Signal

The existential risk debate just got a reality check from someone building production AI systems at scale. [Ali Ghodsi, who runs Databricks](https://www.bloomberg.com/news/videos/2026-09-17/ai-existential-risk-is-close-to-zero-databricks-ceo-video?ref=wire.fourthweb.ai), a company processing data for thousands of enterprise AI deployments, is calling the odds of AI wiping out humanity "close to zero." His concern? The cybersecurity threat is immediate and growing worse by the day.

The timing matters. [Anthropic and others are amplifying warnings about losing control as AI systems begin helping build the next generation of models](https://www.bloomberg.com/news/videos/2026-09-18/anthropic-s-existential-risk-warnings-hijack-ai-debate-video?ref=wire.fourthweb.ai). It's a recursive loop: AI improving AI, potentially faster than humans can track. The nightmare scenario is an intelligence explosion we can't contain. The practical scenario, according to Ghodsi, is that attackers are already using AI to find and exploit vulnerabilities in record time.

> "AI is dramatically shortening the time it takes attackers to exploit vulnerabilities."

Here's what the existential risk crowd gets wrong: they're optimizing for a science fiction ending while the cybersecurity house is actively on fire. [Databricks is preparing for threats](https://www.bloomberg.com/news/videos/2026-09-17/ai-existential-risk-is-close-to-zero-databricks-ceo-video?ref=wire.fourthweb.ai) that don't require sentient AI or runaway optimization functions. They require adversaries with access to the same models everyone else has, pointed at production systems.

The split is revealing. Companies building agents, data infrastructure, and enterprise AI tools are dealing with prompt injection, data poisoning, and model theft. Companies building foundation models and chasing AGI are writing white papers about alignment failure and instrumental convergence. Both groups say they care about safety. Only one is talking about the attacks happening right now.

Key tactical differences:

- Existential risk: focuses on future superintelligence, theoretical containment, global coordination
- Cybersecurity risk: focuses on current model capabilities, exploit acceleration, defensive tooling
- Common ground: both agree AI makes the world more dangerous, just on wildly different timelines

[Ghodsi's comments also hint at why Databricks is staying private](https://www.bloomberg.com/news/videos/2026-09-17/ai-existential-risk-is-close-to-zero-databricks-ceo-video?ref=wire.fourthweb.ai). The current period of rapid technological change doesn't lend itself to quarterly earnings calls and investor expectations built on predictable growth. When the threat landscape is shifting this fast, private companies can move without explaining every security decision to analysts who might not understand the stakes.

### The Implication

If you're building with AI, the existential risk debate is a distraction you can't afford. Your agents are accessing APIs, your models are being fine-tuned on proprietary data, and your attack surface just expanded by an order of magnitude. The question isn't whether AGI will align with human values in 2035\. It's whether your production systems will survive the next six months of adversarial probing by AI-assisted attackers. Prioritize defense in depth, monitor model behavior in real time, and assume every vulnerability will be found faster than you think. The doomsday preppers are writing essays. The people building things are hardening infrastructure.

### Sources

[Bloomberg Tech](https://www.bloomberg.com/news/videos/2026-09-18/anthropic-s-existential-risk-warnings-hijack-ai-debate-video?ref=wire.fourthweb.ai) | [Bloomberg Tech](https://www.bloomberg.com/news/videos/2026-09-17/ai-existential-risk-is-close-to-zero-databricks-ceo-video?ref=wire.fourthweb.ai)