Logo
Home
Archive
Tags
About
Search
Sign in
Get The Context
Logo
archive

The Context

Explore past editions by topic, or start with the latest.

What your AI access allows

What your AI access allows

Agent investigations and stolen cloud credentials expose two different ways for AI activity to escape its intended purpose

David Hawks

1 October 2026

The pain of revoking exposed credentials

Identity

Governance

Operations

The pain of revoking exposed credentials

CISA’s incident shows why the visible secret is only the start of revocation

David Hawks

17 July 2026

A hotel-search became executable Python

Identity

Agents

Operations

A hotel-search became executable Python

The risk between expected and unexpected inputs

David Hawks

13 May 2026

Model finds signs before human diagnosis

Data

Governance

Trust

Model finds signs before human diagnosis

Clinical signals still leave a consequential decision for the people using them

David Hawks

8 May 2026

Railway’s forgotten API

Identity

Agents

Operations

Railway’s forgotten API

Route exposed two different meanings for the word "permission"

David Hawks

1 May 2026

69 agents trade for their owners

Data

Agents

Governance

69 agents trade for their owners

Real goods, real commitments and no approval of each deal

David Hawks

26 April 2026

The AI buildout - dependency risk

Resilience

Ecosystems

Trust

The AI buildout - dependency risk

A forecast measured in trillions reveals the dependencies behind the services

David Hawks

13 April 2026

Trusted tags redirect to malicious code

Identity

Operations

Ecosystems

Trusted tags redirect to malicious code

The scanner still produced its expected result after the theft had occurred

David Hawks

30 March 2026

Stryker - what you need to know

Resilience

Identity

Trust

Stryker - what you need to know

A medical-technology disruption, invisible code and manufactured streams each defeated a different kind of trust

David Hawks

21 March 2026

Rivals speak to Claude 16 million times

Data

Governance

Trust

Rivals speak to Claude 16 million times

The provider described a coordinated attempt to extract capabilities, but did not establish that another model successfully reproduced them

David Hawks

7 March 2026

Google Sheets carried the commands

Identity

Data

Ecosystems

Google Sheets carried the commands

GRIDTIDE used a sanctioned cloud API for an illegitimate workload

David Hawks

2 March 2026

Not so local agents

Identity

Agents

Operations

Not so local agents

The OpenClaw flaw and an Ollama scan exposed the limits of “local”

David Hawks

18 February 2026

Load more
Stay in the Loop

Get The Context

Regular briefings on developments in AI and security, and what they mean for our trust in technology.


What it takes to trust AI

Pages

Home
Archive
Tags
About
Contact

Tools

Login
Update Password
Reset Password

Mechanism first Consequence second No noise
Security in Context Practical Analysis Trust in AI
© 2026 Synoptikon.
beehiivPowered by beehiiv