AI Security 2025 Wrap: 9 Predictions Hit & The AI Bubble Burst of 2026

AI Security 2025 Wrap: 9 Predictions Hit & The AI Bubble Burst of 2026

by Ashish Rajan and Caleb Sima

Humanities, Social Sciences & Thought Computing & Cybersecurity UComputing and Information TechnologyURComputer security
🎙 Ashish Rajan and Caleb Sima 👥 20K 📅 December 19, 2025 ⏱ 63 min 👁 22K 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

AI predictionsAI securitySOC automationAI bubbledata security

Summary

In this season finale, hosts Ashish Rajan and Caleb Sima review their 2025 AI predictions, claiming a 9-for-9 accuracy. They discuss the state of AI utility, noting significant progress in consumer and individual use, but highlight that mature AI production systems remain limited due to cost, reliability, and scalability issues. They confirm that SOC automation became the most tangible real-world AI impact, while AI red teaming also gained traction. They revisit predictions on AI in browsers and operating systems, noting slower adoption due to security concerns like prompt injection. Data security and geo-locking are identified as winners, and they reaffirm that agentic AI was overhyped. Looking to 2026, they predict the bursting of the AI bubble, the rise of self-fine-tuning models, and the controversial idea that ‘AI Engineer’ is just a rebrand for data scientists. They also speculate on the potential disappearance of OpenAI. The conversation is informal and opinion-driven, with a focus on industry trends rather than rigorous analysis.

160 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the information lies in the hosts’ direct industry experience and their ability to synthesize trends from their work with enterprises. They provide concrete examples, such as the cost of AI red teaming products and the evolution of coding with AI. However, the argumentation is largely anecdotal and lacks empirical evidence or citations. The hosts make bold claims, like the AI bubble bursting, but do not provide a structured analysis or data to support these predictions. The discussion is engaging but would benefit from more rigorous evidence.

Scientific Rigor, Source Quality, Title Accuracy

The hosts do not cite specific sources during the episode, and the description only provides links to their own website, newsletter, and LinkedIn. This limits the verifiability of their claims. The title accurately reflects the content, as the episode is a review of predictions and a look ahead. The hosts’ credibility as security professionals adds some weight, but the lack of external references reduces the scientific rigor. No comments were provided for analysis.

177 words

Title / Content Match

The title accurately reflects the content: a year-end review of 2025 predictions and a look ahead to 2026, including the prediction of an AI bubble burst.

Quality & Reliability

7/10

The hosts are experienced security professionals, and the discussion is grounded in their practical observations and industry trends. However, the content is largely opinion-based and lacks formal citations or empirical data, which limits its scientific rigor.

Chapters

Cited Sources

Concurring Sources

Dissenting Sources

  • Gartner Hype Cycle for Artificial Intelligence

Contribution & Novelties

The episode provides a retrospective on AI security predictions, offering insights into what materialized in 2025. The hosts’ claim of 9-for-9 accuracy is notable, though it is based on their own assessment. The discussion on the cost barriers to AI production and the prediction of an AI bubble burst offer a contrarian perspective. The idea that ‘AI Engineer’ is a rebrand of data scientists is provocative and may spark debate.

Pour aller plus loin :

  • AI bubble — Context on the concept of an AI bubble and its potential economic implications.
  • Prompt injection — Overview of prompt injection attacks, a key security concern discussed in the episode.
  • EU AI Act — Information on the EU’s regulatory framework for AI, relevant to the discussion on data security and compliance.

128 words

Radar Profile

The radar profile shows moderate scores across all dimensions, indicating a balanced but not exceptional performance. The highest score is in information quantity, reflecting the breadth of topics covered, while reliability is lower due to the lack of citations and empirical evidence.

Reliability 6/10