CANADA Trends and Developments Contributed by: Kevin West, Andrea Hill, Priya Ratti and Gabriel Potkidis, SkyLaw
How Deals get Done With AI Negotiations have also evolved. Increasingly, AI is being used directly in negotiations. For example, many non-disclosure agreements are now being negotiated AI to AI, with limited human involvement. Participants are also using AI to determine what is the market standard for certain provisions, which can lead to challenges given the potential inaccuracies, particularly with private transactions where publicly available data is limited. It is not uncommon to receive comments on transaction documents that have quite obviously been identified by AI and not properly vet - ted by a human. While AI in many ways makes things more efficient, the tendency of AI to “over-lawyer” the review of deal documents can significantly slow negotiations. Practitioners with real experience and judgement are still required at every stage. AI and data issues are heavily negotiated provisions in technology‑adjacent deals: data rights, cybersecurity procedure, model governance, IP provenance (includ - ing open‑source use), and AI‑driven product claims. Diligence & Valuation For AI‑focused targets, key diligence questions include: • who owns or controls training data; • what contractual and privacy restrictions apply; • how portable the model is; • whether performance is repeatable across custom - ers; and • whether security constraints limit scaling. It can be a challenge to determine a valuation for a company that has financial information only on its legacy businesses (the business built before a new strategic shift, such as adopting AI). Potential acqui - rors will need to make assumptions about the speed and ability of AI implementation to lower costs and increase revenue. Valuing a company with a legacy business using a multiple of historical earnings before interest, taxes, depreciation, and amortisation (EBIT - DA) may not be as meaningful as it once was. Increasingly, acquirors are concerned about deal risks such as government policy and tax changes, multi-
jurisdictional data and privacy regulations, and the ability to recruit and retain key talent. The Challenge of AI Regulation There is little doubt that AI is one of the most trans - formative technologies ever created. Sir Demis Has - sabis, a Nobel winner and co-founder of Google’s DeepMind, has famously stated that AI will be “at least as big as the Industrial Revolution, possibly bigger”, and he draws parallels between AI and “the advent of electricity or even fire”. AI can speed exponentially the discovery of new things such as medicines and make vastly more effi - cient labour-intensive processes such as manufactur - ing and mining. However, many tech giants believe that AI is a fundamental risk to the existence of human civilisation if developed without effective safeguards. Geoffrey Hinton, widely regarded as one of the god - fathers of AI, has sounded the alarm about AI and predicts there is a 10% to 20% chance that AI could lead to human extinction: “These things are getting smarter than us. I’ve come to the conclusion that it’s as though we’re little children playing with a bomb.” Governments considering the risks of AI face a dif - ficult balancing act. Limiting the use of AI could limit the ability to scale and compete with other countries. Anthropic, a large US AI company best known for building the Claude family of large language models (LLMs), reportedly wanted to reduce the possibility its LLMs would be used by the US government for mass domestic surveillance and autonomous weapons. The US government designated Anthropic as a “supply- chain risk to national security” because of its refusal to allow the Department of War to have broader usage rights, signalling that the US is unlikely to regulate AI in the way some would like. Early in the Russia–Ukraine war, many governments and technologists emphasised the need for mean - ingful human control over AI-enabled targeting. Since then, conflicts have increasingly incorporated AI for intelligence analysis and strike planning, and the bat - tlefield use of drones at scale has accelerated, raising concerns that practical guardrails are eroding.
284 CHAMBERS.COM
Powered by FlippingBook