Over the last six months, artificial intelligence has evolved from an essential part of modern business to a potential existential threat, leading many AI leaders to call for the development of the technology to slow down.
The alternative, apparently, is an internet overrun with AI agents and hundreds of billions in losses, at least according to Anthropic boss Dario Amodei.
“I have become convinced that fully addressing the risks requires even more prudence – not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up,” Amodei said in a recent essay called We Must Pace the Frontier.
“We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.”
Left unchecked, Amodei warned, we may lose the ability to understand not only how these systems work, but how to control them. Add to this that governments, such as Australia’s, are still grappling with AI regulation and standards; so, what can businesses do if their essential tools and agents change or become obsolete or restricted?
“If the AI product is updated, the provider’s terms shift, or regulation catches up with how the tool works, you may find the platform you’ve built critical processes around is no longer fit for purpose,” Lauren McKee (pictured), practice leader at commercial law firm LegalVision, said.
According to McKee, Australian businesses need to be checking “whether the provider can use your prompts or outputs to train their models, how they handle confidentiality and data security, whether they actually comply with privacy laws, what service levels they are committing to, and whether they are trying to unreasonably limit their liability if something goes wrong”.
One option, McKee said, is to look at possible “exit rights”.
“If the product changes or stops working for your business, you need to be able to retrieve your data and transition without being locked in,” she said.
“Useful protections include service level commitments, notice of material changes, restrictions on model changes that affect performance, data portability, prompt and output ownership terms, confidentiality and privacy obligations, cyber incident notification, business continuity commitments and meaningful termination rights.
“Liability caps should not quietly exclude the very losses most likely to occur, such as privacy breaches, IP claims or service disruption.”
AI tools, McKee contended, need to be treated like any other critical supplier, and documented fallback options should already be locked in as a matter of course in case of disruption.
“Organisations can reduce reliance by designing their AI use around business processes, not one provider’s product,” McKee said.
“SMEs should also set approval rules so staff cannot independently build critical workflows on unapproved tools.”
All the usual rules of basic governance should apply to AI, from business policies, tool registers, risk assessments, human reviews, and incident reporting. With all that in place, McKee believes any business should be able to take the loss or disruption of any AI tool in stride.
“AI access can change quickly, and businesses should not build essential operations on assumptions they do not control,” she said.
Want to see more stories from trusted news sources?Make Cyber Daily a preferred news source on Google.
