Andrei Baciu on the evidence dilemma, the pacing problem, and why parliaments should copy each other
BY BEATRIZ REY
On October 22, at the Inter-Parliamentary Union (IPU)’s webinar “Extreme risks from AI: the parliamentary role in setting limits,” most of the discussion focused on how lawmakers should regulate AI. But one speaker, Andrei Baciu, a member of Romania’s Chamber of Deputies, turned instead to how parliaments themselves can build the capacity to deal with the technology.
Baciu told participants that Romania’s Parliament is about to train its members and staff on AI. From November 10 to 12, two international experts, one based in New York and one in Singapore, will run a three-day program in Bucharest for members and staff of both the Chamber of Deputies and the Senate. The training, which Baciu proposed and is organizing with the United Nations and the United Nations Development Programme (UNDP), will cover the fundamentals of AI, its opportunities and risks, ethics, governance and regulation, and the role of parliaments in dealing with emerging technologies. It grew out of conversations Baciu began earlier this year with UN officials in New York and Geneva.
Baciu has also formally proposed a dedicated parliamentary committee on AI, as a space for dialogue and cooperation on regulation, ethics, emerging risks and AI’s effects on society and public institutions. The proposal has been submitted and now has to go through parliamentary procedure; the committee does not yet exist. (Romania’s Senate has already added AI to the remit of its communications and IT committee.)
His reasoning starts with what is known as the “evidence dilemma.” Parliaments must make decisions about AI before there is solid evidence to back them, and those decisions are made by political majorities, not by whoever has the best argument. If members don’t understand the technology, they can’t weigh the evidence that does exist, and no expert will be able to persuade them.
His answer is not to turn MPs into specialists. The technology changes too quickly for that: someone who becomes an expert in agentic AI today will be behind by the time the next development arrives. The goal is broader literacy and some fluency, enough for a parliament to know what questions to ask and to make good use of independent experts when it brings them in.
It also matters that staff are being trained alongside Members. In most parliaments, the people who brief Members, draft amendments and manage committee work are the ones who will actually use AI tools and judge their output. A parliament that trains only its Members, or only its IT staff, will struggle to adopt the technology well.
Baciu also described a national public consultation on AI, open from August to mid-October, asking Romanians about their priorities and concerns, with a parallel track for business, academia, and the public sector. Results are expected at the end of November. His argument is that society now moves faster than institutions can adapt, what POPVOX Foundation calls the “pacing problem,” and that this gap erodes public trust. Using digital tools to hear directly from citizens is one way for parliaments to narrow it.
He closed with a practical point. A parliament that starts from scratch, with consultations, expert input and drafting, can easily spend a year before it has anything to show. Borrowing a practice that already works elsewhere is much faster. That is why he called for a standing network where parliaments share what they are doing, and pointed colleagues to the IPU’s inventory of parliamentary AI initiatives as a place to start.
I will follow up on Romania’s experience after the training ends on November 12, and again when the consultation results come out later this month.
Modern Parliament (“ModParl”) is a newsletter from POPVOX Foundation that provides insights into the evolution of legislative institutions worldwide. Learn more and subscribe at modparl.substack.com.
