AI in Environmental Assessments: Experts Sound Alarm Over “Robodebt-Style” Risks to Biodiversity
Sydney, Australia – A proposal by Australia’s mining industry to leverage Artificial Intelligence (AI) for expedited environmental permit approvals is facing strong opposition from scientists and conservationists, who warn of a potential collapse in regulatory oversight akin to the notorious “Robodebt” scandal, with devastating consequences for endangered species.
Concerns Over Ambiguous Legislation and Data Gaps
The Australian Biodiversity Council, an independent group of experts from 11 universities, has voiced significant concerns, arguing that Australia’s current environmental laws are ill-equipped for AI-driven decision-making. “Vague environmental laws are not suitable for AI,” stated the council.
While acknowledging AI’s potential for simpler tasks, the council cautions against automating complex environmental impact assessments. “This could lead to Robodebt-style failures, where computers make flawed decisions without transparency,” they warned, underscoring the risk of pushing vulnerable species closer to extinction. The Robodebt scheme, a failed automated debt recovery program, wrongly accused hundreds of thousands of welfare recipients between 2015 and 2019, leading to widespread administrative errors and a significant public scandal.
The Perils of Vague Regulations and Insufficient Data
Liz Ashbey, Policy and Innovation Lead at the Biodiversity Council, highlighted that Australia’s primary environmental legislation is “riddled with vague language and broad ministerial discretion.” She explained that such ambiguity already complicates human-led assessments and would be even more problematic for AI. Establishing clear guidelines within national environmental standards, even without AI, is crucial for streamlining assessments and essential for any future AI integration.
Professor David Lindenmayer, a forest ecologist at the Australian National University and a member of the council, pointed to a critical data deficit. He noted that research indicates a third of Australia’s endangered species remain unmonitored, with others possessing only fragmented data. Human assessors bridge these gaps through expert consultation. “AI decisions are only as good as the data they rely on, and good data is unavailable for most of Australia’s endangered species – often not even basic location data exists,” Professor Lindenmayer explained. Automating these assessments with AI risks decisions based on erroneous or outdated information, undermining biodiversity protection efforts.
Professor Hugh Possingham, a leading conservation biologist from the University of Queensland, added that AI tools typically require extensive training data. He argued that the last “20 years of EPBC Act approvals are not suitable for this,” as the law has demonstrably failed to adequately protect the environment.
Industry and Government Responses
Tanya Constable, CEO of the Minerals Council of Australia, described the comparison to Robodebt as “disappointing,” asserting that the proposed innovative approach could enhance environmental protection while improving efficiency. She emphasized that the proposed approach “supports human decision-making with AI tools.”
A spokesperson for the Australian Federal Government indicated that budgetary decisions would be made “in due course.” However, the Department of the Environment is reportedly exploring ways to simplify the application process using AI. The spokesperson reiterated that “decisions on whether to approve or reject projects should always be made by assessment officers, not by AI.”
