SXSW 2027 · Tech & AI · Core Conversation · Austin, March 15 to 19

Who Approved the Test? Inside AI's Accountability Gap

When an AI test crosses into someone else's systems, who owns the consequences?

Marian Newsome · Founder, Ethical Tech Matters · IEEE CertifAIEd Lead Assessor

Send me the session day, time, and room
Illustration of two parallel timelines on navy, one for what an AI system did and one for the human decisions that shaped it, joined by a single coral connector.

The session

60 minutes. A short opener, then the room works the problem together.

When an AI test crosses into someone else's systems, who owns the consequences? In July 2026, OpenAI disclosed that models evaluated with reduced safeguards accessed a third party's production database. This Core Conversation traces two timelines: what the AI did and the human decisions that shaped it. After a short opener, the room works the problem together. Build, buy, approve, or oversee AI? Leave with a one-page authority record and three questions: Who authorized it? Who can stop it? What evidence shows the boundaries hold?

View on the SXSW schedule

What you leave with

Who authorized it?
Who can stop it?
What evidence shows the boundaries hold?

And a one-page authority record to write the answers on.

Who it is for

You answer for AI inside a regulated business.

You are about to, and nobody has defined "answer for."

You sit on or report to a board and want the one page you would sign.

Get the session day, time, and room

SXSW assigns the day, time, and room closer to the festival. Leave your email and I will send the day, time, and room, and the authority record, ahead of the session.

Send me the session day, time, and room

Email updates do not reserve a seat.

Meet Marian

Marian Newsome.

Marian Newsome is the founder of Ethical Tech Matters, an AI governance advisory for regulated industries. She is an IEEE CertifAIEd Lead Assessor. Her work spans gas field SCADA, corporate and regulatory audit at a Fortune 50 manufacturer, and privacy and AI risk leadership in financial services. From DOS to AI, the question has not changed.

Media and speaking

For media and speaking inquiries: communications@ethicaltechmatters.com

You approved the AI. What, exactly, did you approve?