menu
{ "item_title" : "Red Team", "item_author" : [" Empire Publishing "], "item_description" : "Companies are shipping AI fasterthan they can secure it.The attack surface here is genuinely new: prompt injection, tool and agent abuse, model supply chain, data walking out through a chatbot that was only meant to answer questions. And the response is hardening into a permanent function - the World Economic Forum's Global Cybersecurity Outlook 2026 found the share of organisations assessing the security of their AI tools rose from 37% to 64% in a single year.Two things decide whether you get hired, and neither is a certificate. First, the job title is a trap: very few postings say 'AI red teamer', while a far larger share of ordinary security roles now carry this work inside the description - search by title and you will conclude the market does not exist. Second, what gets read is evidence: a documented bypass, a contribution to an open testing tool like garak or PyRIT, a write-up of how you broke a model and what you would do about it.So the book is built around producing that evidence rather than passing an exam, and around how the function itself is changing - from one-off project to standing assurance, because the model changes underneath you and because customers and regulators increasingly ask for proof.The craft is anchored in free, citable frameworks rather than vendor marketing: the OWASP Top 10 for LLM Applications and the NIST AI Risk Management Framework. Pay is handled carefully - no government wage series covers this role, so the wilder numbers circulating online are left where they belong.It is written for someone already in security or engineering who wants to specialise upward rather than start over. This is not a first job in technology, and the book says so instead of selling a shortcut. Testing systems you do not own or lack written permission to test is unlawful in most jurisdictions, and authorisation and scope are taught from page one. An Australian appendix covers the local position separately.New attack surface. Very few people who can test it. That is the opening.", "item_img_path" : "https://covers4.booksamillion.com/covers/bam/9/79/819/279/9798192793107_b.jpg", "price_data" : { "retail_price" : "17.99", "online_price" : "17.99", "our_price" : "17.99", "club_price" : "17.99", "savings_pct" : "0", "savings_amt" : "0.00", "club_savings_pct" : "0", "club_savings_amt" : "0.00", "discount_pct" : "10", "store_price" : "" } }
Red Team|Empire Publishing

Red Team : Break Into AI Security - Testing the Models, Agents and Pipelines Companies Are Shipping Faster Than They Can Secure Them (A Field Guide)

local_shippingShip to Me
In Stock.
FREE Shipping for Club Members help

Overview

Companies are shipping AI faster
than they can secure it.

The attack surface here is genuinely new: prompt injection, tool and agent abuse, model supply chain, data walking out through a chatbot that was only meant to answer questions. And the response is hardening into a permanent function - the World Economic Forum's Global Cybersecurity Outlook 2026 found the share of organisations assessing the security of their AI tools rose from 37% to 64% in a single year.

Two things decide whether you get hired, and neither is a certificate. First, the job title is a trap: very few postings say 'AI red teamer', while a far larger share of ordinary security roles now carry this work inside the description - search by title and you will conclude the market does not exist. Second, what gets read is evidence: a documented bypass, a contribution to an open testing tool like garak or PyRIT, a write-up of how you broke a model and what you would do about it.

So the book is built around producing that evidence rather than passing an exam, and around how the function itself is changing - from one-off project to standing assurance, because the model changes underneath you and because customers and regulators increasingly ask for proof.

The craft is anchored in free, citable frameworks rather than vendor marketing: the OWASP Top 10 for LLM Applications and the NIST AI Risk Management Framework. Pay is handled carefully - no government wage series covers this role, so the wilder numbers circulating online are left where they belong.

It is written for someone already in security or engineering who wants to specialise upward rather than start over. This is not a first job in technology, and the book says so instead of selling a shortcut. Testing systems you do not own or lack written permission to test is unlawful in most jurisdictions, and authorisation and scope are taught from page one. An Australian appendix covers the local position separately.

New attack surface. Very few people who can test it. That is the opening.

This item is Non-Returnable

Details

  • ISBN-13: 9798192793107
  • ISBN-10: 9798192793107
  • Publisher: Independently Published
  • Publish Date: August 2026
  • Dimensions: 9 x 6 x 0.52 inches
  • Shipping Weight: 0.74 pounds
  • Page Count: 246

Related Categories

You May Also Like...

    1

BAM Customer Reviews