Government

House Officials Tell MPs the Commons AI Strategy Is Unpublished and U.S. Courts Could Reach Its Data

· 6 min read · Government

House Officials Tell MPs the Commons AI Strategy Is Unpublished and U.S. Courts Could Reach Its Data

The House of Commons has an artificial intelligence strategy covering 2025 to 2027 that has never been published, and its own technology chief says Commons data stored with Microsoft carries risks under the U.S. CLOUD Act, The Hub reported on October 24. Both admissions came on October 20, when Clerk of the House Eric Janse and two senior administrators appeared before the Standing Committee on Procedure and House Affairs for its study of AI and Canada's democratic institutions.

Conservative MP Michael Cooper asked whether the strategy had been made public. Natalie Foster, Clerk Assistant and co-chair of the House AI Working Group, said it is available internally only and agreed to provide it to the committee. Conservative MP Kelly DeRidder then asked whether the Microsoft agreement protects the House from the CLOUD Act. Christopher Henry, Director General of Digital Enablement and Innovation, said contracts require Microsoft to contact the Speaker and the law clerk if data is sought, but that U.S. court orders could reach the information before the House is able to intervene. He added that the House is talking to Canadian AI firms about parallel pilots.

The stakes are practical. The House already runs a Copilot pilot for members chosen by party whips and has used AI to produce a first draft of Hansard for more than a year. A second panel, from the International Association of Conference Interpreters, warned that machine interpretation is not fit for Parliament, citing a World Health Organization evaluation in which only one speech in 90 received a passing grade.

What it means in practice

House Officials Tell MPs the Commons AI Strategy Is Unpublished and U.S. Courts Could Reach Its Data — contextual photo

For Parliament, the question is less whether to use AI than who controls the infrastructure and the rules. An unpublished strategy makes it hard for MPs and the public to see which tools are approved, what data they touch and how errors are caught, in a bilingual institution where interpretation and translation carry legal weight. The data-residency concern echoes debates across Ottawa, from Cohere's push to win federal buyers to Quebec's exploratory deal to test Cohere in its public service. The same committee has already heard that chatbots fed Quebec voters faulty election information.

Officials also conceded limits on detecting machine-generated content. Janse warned the committee could receive thousands of submissions from a single person, and Henry said the House lacks the tools to detect all AI-generated material sent to it, although Foster said abuse has not reached worrying levels. A proposal on AI disclosure for written briefs is going to the Liaison Committee.

What to watch: whether the committee publishes the internal strategy once it is tabled, whether the House moves from guidelines to binding rules, and whether pilots with Canadian AI providers lead to a sovereign option for sensitive parliamentary data.

← Back to AI4Canada