With renewed attempts to introduce
AI regulation in the UK. What does this mean for enterprises, and should
business leaders welcome it?
Throughout history, society
has created and
evolved regulatory policies in line with societal
needs, market opportunities and business trends. But never has there been a
faster-developing and higher-stakes technology than AI. All market regulation -
financial, transport or tech focused - has been somewhat shaped by
the strength and direction of political winds. Given the global nature of AI and the potential to affect every aspect of our personal
lives, society and even the climate, the
case for thoughtfully constructed regulations has even greater consequence.
As
with many aspects of regulation, the speed of technological evolution often
outpaces the ability for regulatory bodies to identify what guardrails need to
be established, which maximise the benefits, whilst minimising the risks. AI is proving no different.
What AI regulation already exists?
Whilst there may not be global
agreement on the objectives of AI regulation and how far rules should extend, there are
still some major regulatory frameworks already in play, such as The EU AI Act, which provides a unified set
of risk based, safety and transparency rules for AI across the EU. China also
has established legally binding regulations such as transparency around
algorithms and, from September this year, AI-generated content must be
labelled.
Other countries are in the process of establishing a federal
regulatory framework (such as Canada’s Artificial Intelligence and Data Act
(AIDA)). Then there is the US, which
is evolving its regulatory position, with the White House recently ordering federal agencies to expand use of AI and
reverse some of the Biden-era safeguards.
What AI regulation does the UK
have?
We have gone from an initial hawkish stance on AI risk, to
favouring a more principles-based approach, which leverages existing laws and regulations.
In 2023 at the AI Safety Summit at Bletchley Park, the UK
government spearheaded conversations around existential threats posed by AI. But it has since
moved away from adopting a centralised AI regulator. Instead, it’s choosing to empower
regulatory bodies to look at the impact of AI on the markets they
regulate. For example, the Bank of England, the Financial Conduct
Authority, the MHRA and the Competitions and Market Authority. The Information
Commissioner’s Office also recently issued draft guidance on using AI.
This
transition of positioning has been further exemplified by renaming the AI
Safety Institute to the AI Security Institute. This was a small but
symbolically significant move indicating that concerns have switched from
algorithmic fairness and bias, to defending against malicious use of AI and
ensuring geopolitical resilience.
Do we need global AI policy alignment?
At
a global level there is a hodgepodge of existing and emerging AI regulations.
Many Governments around the world are still not sure how to frame, manage and
mitigate the risks. Or even what risks to manage. They are caught between
wanting to leverage the power of AI for various capabilities and the potential
economic gold rush (especially amid anaemic economic growth for many
countries); but then worrying about the potential unintended societal impact of
this technology.
But
is misalignment a problem? For enterprises it creates a complex operating
environment with the associated increase in costs of doing business across
jurisdictions. At a societal level, many
argue that AI regulatory alignment could either exasperate or close global
inequalities.
What does this mean for business?
All these moving parts present multinational businesses with a
complex environment to
navigate. Making internal policies,
frameworks and governance important for several
reasons.
Firstly, traditional
approaches such as applying the high watermark, are not enough. This practice
only works when regulations in different jurisdictions are broadly trying to
achieve the same objective. But with AI, the current regulations often have different
objectives. For example, in China AI regulations focuses on ensuring
political and social stability, whereas in the EU, regulations focus on
ensuring human rights and minimising bias.
Corporations therefore need
to be more contextually aware when they decide how to build, deploy and use AI,
remaining abreast of evolving rules in each jurisdiction. They can no longer
simply set their systems to meet the global high watermark and be assured of
meeting regulatory obligations.
Secondly,
having a strong AI governance model can be a
competitive differentiator. Given high consumer alert around potential risks
of AI, enterprises will win trust from their
customers, employees, partners and ecosystem by adopting robust AI governance. Going beyond the minimum
regulations, to thoughtful and considered
policies, a company provides reassurance and clarity.
Apple is a case in point.
It built consumer trust by championing user privacy and positioning itself as a
protector of data. This differentiation contributed to a strong brand value and
customer loyalty for Apple.
Finally, forging a strong AI governance policy frees teams for
creativity. Criticism of policies or regulations
are frequently grounded
in the belief that they slow innovation. But
the opposite is true. It may seem counterintuitive, but greater
ambiguity around the use of technology, invariably
slows adoption and value realisation. This is because employees don’t know the boundaries of
what they can do and can’t do. They end up with analysis-paralysis impacting
operating costs and speed-to-value creation.
Creating an AI governance
framework
Just
like AI’s capabilities are evolving, so too will the enterprise's governance
model. Here
are some suggestions on how companies can lay strong foundations today:
- Start
building trust by sharing your
AI vision, how
you intend to use it and the ethics by which you will govern its use.
- Next, embed your AI ethics deeply into
company culture through training programmes, internal communications and
reinforcement of messaging.
- Build a dynamic and
risk-weighted governance
framework that adapts to the fast-moving pace of
the technology,
the increasing diversity of use-cases and global regulatory
variances. Also include an AI-incident response framework.
- Consider differing
jurisdictional regulations when architecting your AI platforms. Such as
taking a jurisdictional zoning and modular compliance approach.
- Finally, invest in your people; the tools; and regulatory intelligence so that
the right people,
with the right capabilities are on top of the dynamic AI regulatory landscape.
For further
information visit Tarralugo
