Know what you're signing up for

Independent risk ratings for the AI products people are actually using — scored on privacy, bias, job threat, autonomy, and lock-in.

12 products rated Last updated see what changed No vendor money
F 1 D 7 C 3 B 1
Latest rating Search

Perplexity

Now ships an agent that works your files and apps around the clock, while still being sued by nearly every major newsroom.

What changed Grade C → D risk 5.4 → 6.2(Worse)

Perplexity spent 2026 building Computer, an agent with standing access to a user's files and apps: a Mac version shipped in April, a $20-a-month Pro tier opened in May, and a local, on-device version called Portable Computer reached Windows RTX PCs on September 14. The company now credits that product line with most of its revenue growth, and its CEO has said the goal is to run businesses as autonomously as possible while dismissing the resulting job losses as an acceptable trade. Autonomy moved from 6 to 8, job threat from 6 to 7, and lock-in from 3 to 4, taking the grade from C to D.

D
High Risk
Privacy
7
Bias
5
Job threat
7
Autonomy
8
Lock-in
4

The Index

Highest risk first
F
Autonomous Tesla New

Tesla Autopilot / FSD

A $15,000 promise of self-driving that's killed people and still calls itself 'Full Self-Driving.'

Grade F held risk 7.8 → 8.2(Worse)
high risk
D
Creative Midjourney, Inc. New

Midjourney

Your prompts and pictures are public unless you pay $60 a month, and three Hollywood studios are asking a court to switch it off.

high risk
D
Enterprise Otter.ai, Inc. New

Otter.ai

The bot your colleague invited to the call, recording everyone else on a consent nobody but the host gave.

high risk
D
Chatbot Google New

Google Gemini

Google already had all your data. Now it has a chatbot that can actually use it.

Grade D held risk 7.2 → 7.4(Worse)
high risk
D
Chatbot OpenAI

ChatGPT

Knows more about you than your therapist. Reports to a company that wants to be God.

Grade D held risk 6.6 → 7.4(Worse)
high risk
D
Enterprise Workday

Workday AI Recruiting

The screening layer between you and a billion job applications, with its bias testing sealed under attorney-client privilege.

high risk
D
Surveillance Amazon

Amazon Alexa

A corporate spy you bought with your own money, placed in your bedroom, and thanked for the weather report.

Grade D held risk 6.2 → 6.6(Worse)
high risk
C
Assistant Microsoft New

GitHub Copilot

Your AI pair programmer that learned to code by reading everyone else's code without asking.

Grade C held risk 5.2 → 5.8(Worse)
use with caution
C
Hardware Meta / EssilorLuxottica

Ray-Ban Meta Smart Glasses

A camera and microphone worn on your face, wired to Meta. The people around you never agreed to any of it.

use with caution
C
Creative Adobe

Adobe Firefly

Trained on data it actually paid for, which is rarer than it should be. Getting back out of Adobe is the expensive part.

use with caution
B
Surveillance Microsoft New

Microsoft Recall

A screenshot of your screen every five seconds, indexed and searchable. It never leaves the laptop, which is the only reason this is not an F.

moderate risk

Don't Miss the Next Warning

Get notified when we publish new cautionary tales where AI turns sinister. Join our community of readers exploring the dark side of artificial intelligence through classic literary styles.

No spam, ever
Unsubscribe anytime
Weekly digest

By subscribing, you agree to our Privacy Policy, Terms of Service, and Cookies Policy. We respect your privacy and will never share your email address.