Bring back technology assessment

In 1995 the US Congress shut down one of its smallest agencies. The Office of Technology Assessment had a staff of under two hundred and a budget of around $20 million a year, a rounding error in federal terms. Over two decades it had produced several hundred reports on subjects from acid rain to missile defense to genetic testing. Then the new majority elected in 1994, looking for visible savings and skeptical of in-house expertise, cut its funding. It closed that September.

More than thirty years later, I think that decision looks worse every year. This is an opinion piece, so I'll say plainly what I believe: democracies need public, independent bodies whose job is to look ahead at new technologies and tell legislators and citizens what they might do, to whom, and what the options are. Artificial intelligence and synthetic biology make the case more urgent, not less. But the arguments against are serious enough to deserve a proper hearing, and I'll try to give them one.

What OTA actually was

OTA was created by the Technology Assessment Act of 1972 and began work in 1974. It served Congress, not the executive branch, and was overseen by a board of six senators and six representatives split evenly between the parties. That structure mattered. Requests came from committees of both parties, and reports were reviewed by outside experts before publication.

The reports did not usually tell Congress what to do. They laid out options and their likely consequences. Some made powerful enemies: OTA's analyses of the Strategic Defense Initiative in the 1980s raised doubts about whether the program's goals were technically achievable, which did not endear the office to the program's supporters. Others were simply useful, unglamorous work that helped legislators understand subjects they were about to legislate on.

Its closure did not end the idea in Washington. The Government Accountability Office began doing technology assessments in the early 2000s, and in 2019 it set up a dedicated science, technology assessment and analytics team. The Congressional Research Service and the National Academies also do valuable work. Bills to revive OTA itself have been introduced repeatedly over the years. None has passed.

What Europe kept

Europe went the other way. The European Parliamentary Technology Assessment network, founded in 1990, now links more than a dozen parliamentary and national bodies. Britain's Parliamentary Office of Science and Technology dates from 1989. The German Bundestag has had its own Office of Technology Assessment since 1990. The Netherlands has the Rathenau Institute, and the European Parliament its Science and Technology Options Assessment panel.

Europe also experimented with something OTA mostly didn't: asking citizens. In 1987 the Danish Board of Technology held a consensus conference on gene technology in industry and agriculture. A panel of lay people, chosen to reflect a range of backgrounds, studied the subject, questioned experts in public and wrote its own recommendations for parliament. The format was copied in many countries over the following decades, from the Netherlands and Britain to Japan and South Korea. The Danish board later ran World Wide Views, which held simultaneous citizen deliberations in dozens of countries on climate change and biodiversity.

These institutions have their own troubles. The Danish board lost its state funding in 2012 and survives as a foundation. Several offices are small and dependent on the goodwill of parliaments. But they exist, and when a parliament wants a careful account of a technology, someone is already on staff.

Why AI and synthetic biology change the argument

Three features of today's emerging technologies make the case for independent assessment stronger than it was in OTA's era.

Expertise is concentrated in companies. In the 1970s much of the relevant knowledge about nuclear power or pesticides sat in universities and government labs. Today the most capable AI systems are built inside a handful of private companies, and their workings are understood in detail mainly by their own staff. Governments increasingly rely on what those companies choose to disclose. Synthetic biology is less concentrated, but its commercial frontier is still largely private.

Assessment has narrowed to security. Where governments have built new capacity, it has tended to focus on catastrophic or national-security risks. Britain renamed its AI Safety Institute the AI Security Institute in 2025; the US equivalent was reorganized with a similar emphasis the same year. Those are legitimate concerns. But they leave out labor markets, competition, the environmental footprint of data centers, effects on education and public information, and who captures the gains. Synthetic biology saw the same narrowing two decades ago, when its governance debate shrank to biosecurity screening and left the economic questions to no one.

The technologies are converging. AI tools now help design proteins and genetic sequences; the 2024 Nobel Prize in Chemistry recognized exactly that. Assessments organized one technology at a time, by separate agencies, will keep missing the interactions. The progress traced in our look back at synthetic biology's early promises came largely from unexpected combinations, and so will the risks.

The international layer

Some of these technologies cannot be assessed by national parliaments alone. Solar geoengineering is the obvious case: a technique whose effects would cross every border, debated mostly in a few rich countries. Gene drives, designed to spread through wild populations, raise similar issues.

Proposals for technology assessment at UN level are not new. Civil society groups, ETC Group among them, pressed for an international technology assessment mechanism in the run-up to the Rio+20 summit in 2012, a moment discussed in our piece on that conference. The 2030 Agenda in 2015 created a Technology Facilitation Mechanism, though its focus is on spreading technology rather than scrutinizing it. The Convention on Biological Diversity has set up a horizon-scanning process for synthetic biology. And in August 2025, following the Global Digital Compact agreed the previous year, the UN General Assembly established an independent international scientific panel on AI and a global dialogue on AI governance.

Each of these is narrow, slow or both. But together they show that governments accept, at least in principle, that some technologies need shared assessment. The gap is a standing, general capacity that can look at technologies before they become crises, with governments from the global South involved from the start rather than consulted at the end.

The counter-arguments, taken seriously

"Assessment is too slow for fast technologies." This is the strongest objection. David Collingridge described the problem in 1980: early on, a technology is easy to change but its effects are hard to foresee; by the time effects are clear, it is entrenched. A two-year study of AI models will describe systems that are already obsolete. The answer is not to give up but to change the product. Good assessment in fast fields looks less like a definitive report and more like continuous horizon scanning, short briefings, and scenarios about underlying dynamics – concentration, dependency, labor effects – that change more slowly than any single product.

"It will be captured or politicized." OTA was accused of partisanship, and any assessment body can be captured by the industries it studies or the politicians who fund it. That is a real risk. It is also an argument for the design features OTA got right: bipartisan oversight, external peer review, published methods, and a mandate to present options rather than verdicts.

"We already have this." The US has GAO's assessment team, the National Academies and the Congressional Research Service. Europe has its network. The UN has its panels. Fair enough, as far as it goes. But much of this capacity is small, reactive or tied to particular technologies. What's missing is not any assessment at all, but enough of it, with a broad enough mandate, carried out independently of the companies being assessed.

"Citizen panels are theater." Participatory exercises can be designed to produce consensus for decisions already made. Some have been. But the best consensus conferences changed how parliaments understood issues, because lay panels asked questions experts had not thought to ask. The fix is to commit in advance to how their recommendations will be considered.

"It will slow innovation." Assessment that informs rather than decides should not, in itself, stop anything. Societies that understand a technology may adopt it with more confidence and fewer backlashes. The years of public conflict over GM crops are a reminder of what happens when that understanding is skipped.

What a revived version should look like

If I were drafting the mandate, I would ask for a few things:

• Independence in law, with multi-year funding that cannot be cut in a single budget cycle.
• A broad remit covering economic, social and environmental effects, not only safety and security.
• Speed options: quick briefings alongside longer studies.
• Participation built in, including citizen panels whose outputs carry formal weight.
• Access to information, including the ability to request data from developers of high-impact systems under confidentiality.
• International links, so that national bodies feed into, and draw on, a shared global capacity that includes the South.

None of this is radical. Most of it existed, in some form, in an agency Congress closed to save around $20 million a year. The technologies arriving now, from gene drives to general-purpose AI, are far more powerful than anything OTA studied. Choosing to face them without that kind of public foresight is a decision too, and it is the one we are currently making.