Representative August Pfluger introduced the Reliable Artificial Intelligence Research Act on 30 September 2026, a House bill that would put the Department of Homeland Security (DHS) in charge of two national AI security research competitions, San Angelo Live reported. The proposed programme would invest $10 million over five years in work on AI interpretability and adversarial robustness.
Key points
- DHS would run one competition for interpretability research and another for adversarial robustness research.
- The bill proposes $10 million for the programme over five years.
- Pfluger’s bill is a House companion to legislation introduced by Senators Jim Banks and Maggie Hassan.
DHS would run two research competitions
The competitions would reward research in two related areas. Interpretability concerns how an AI system arrives at a decision, with the aim of making its behaviour more transparent and predictable. Adversarial robustness concerns the system’s resistance to malicious manipulation. Under the proposal, DHS would administer the competitions and distribute rewards for research in those fields.
Pfluger cast the bill as a national security measure. He said AI was becoming more deeply embedded in the economy and national defence, and argued that adversaries were looking for ways to manipulate it. In his account, the competitions would put researchers to work on systems that are “more transparent, more predictable, and more resistant to attacks”.
The two subjects give the department distinct research tasks. One asks researchers to examine how systems produce decisions. The other asks them to make those systems harder to manipulate. Pfluger’s stated aim connects both to the use of AI in settings where an exploited vulnerability could matter beyond the system itself.
The proposed programme would invest $10 million
The legislation would invest $10 million over five years in the DHS programme. The competitions are intended to encourage work across industry, universities and independent research. That approach would direct federal funding towards research on the behaviour and defences of AI systems, with the department responsible for running the contests.
Americans for Responsible Innovation supports the bill. Its president, Brad Carson, said AI models were being deployed in vulnerable systems before their decision-making was fully understood. He warned that an incident could have effects across government agencies and argued that investment in assurance and interpretability researchers would help the country “get ahead of crises before they occur”. Those are Carson’s reasons for backing the proposal, rather than findings produced by the competitions.
Carson’s concern centres on decisions made inside deployed models. The bill pairs research into those decisions with research into attacks on systems, making both subjects part of the same proposed DHS programme. Its five-year funding proposal would support competitions in those areas rather than a single research project.
Banks and Hassan introduced the Senate counterpart
Pfluger’s bill is a House companion to bipartisan Senate legislation introduced by Senator Jim Banks, a Republican from Indiana, and Senator Maggie Hassan, a Democrat from New Hampshire, San Angelo Live reported. The Senate legislation was developed in coordination with technology industry stakeholders.
