AI in the Workplace: Responsible Use for Businesses

Artificial intelligence offers huge growth potential for businesses – and equivalent dangers. Read our 10 Commandments for AI Safety in the Workplace and then download our Company AI Policy Template.

Download the AI Policy Template Now Microsoft Data & AI

The AI Dilemma

Since the arrival of ChatGPT in November 2022, organisations have sought to bring the benefits of AI to the workplace. Tech giants such as Microsoft and OpenAI as well as smaller players have launched products; and businesses are encouraged to get on board – or get left behind. More recently, however, many are realising AI is not without dangers. Getting the right balance between risk and reward is something we must all try to achieve.

What Do We Mean By AI, Anyway?

The UK government’s National Cyber Security Centre (NCSC) says AI is, ‘Any computer system that can perform tasks usually requiring human intelligence. This could include visual perception, text generation, speech recognition or translation between languages.’

As Intersys’ MD Matthew Geyman says, ‘What people are mostly talking about since ChatGPT and equivalents emerged is generative AI, which creates new content based on complex data patterns and includes large language models (LLMs).’

The Benefits of AI

LLMs can be prompted to draft emails, reports and articles; summarise documents and extract key insights; generate images; translate text between languages; schedule and organise calendars; brainstorm ideas – this list is only scratching the surface. As many of us have discovered, AI in the workplace can be a huge time-saver and create valuable ‘first draft’ information in seconds, which we can then oversee and refine with human insight.

Which AI Platform?

In a fast-moving situation, AI is being rolled out across many platforms. These include free and upgradeable paid-for services such as ChatGPT and ClaudeAI, AI integrations into software such as Copilot for Microsoft 365 and even AI responses in search engine queries. We are nearly all becoming users of AI in the workplace in one form or another. However, without proper safeguards, this AI use can pose a significant danger.

Dangers of AI 1

‘Lies, Damn Lies… and AI’

AI frequently makes confident assertions that are partly or wholly inaccurate. It even fabricates links and citations. Research from the Tow Center for Digital Journalism in the US asked generative AI search tools to retrieve and cite news content: they provided inaccurate answers in 60 percent of queries. Meanwhile, in the UK, the High Court has instructed senior lawyers to take urgent action to prevent the misuse of AI after dozens of case-law citations presented at court were found to be fictitious.

Dangers of AI 2

‘Sure, Let Me Steal that Data for You’

Malicious users manipulate AI to encourage toxic content or even to reveal confidential information. These are known as ‘Prompt Injection Attacks’. Furthermore, ‘Data Poisoning Attacks’ tamper with the data an AI model is trained on to produce undesirable outcomes – heavily biased outputs or even security breaches. The point here is that AI is essentially ‘gullible’; it can be manipulated to criminal ends. These kinds of attacks are only expected to increase.

Dangers of AI 3

‘Whoops, We Released Everyone’s Salary’

Some AI models such as Copilot for Microsoft 365 are deeply integrated with company data. This is useful: they can cross-reference meetings, notes and spreadsheets to instantly create, say, a PowerPoint presentation. However, if you’re integrating AI into your processes, you must ensure you properly assign access privileges to users. Otherwise, AI may reference material an employee should never have clearance to see and potentially return salary information, staff appraisals, intellectual property… you name it.

Similarly, shared ChatGPT chats – conversations you can share with others via a unique link - have been shown to be publicly searchable, revealing highly sensitive information. Tread with care.

AI in the Workplace: What’s the Verdict?

Providing you introduce proper safeguards – and understand its limits – AI may be highly useful to your organisation. If you don’t, it could be disastrous. Look at it this way: would you consider using the internet today without appropriate firewalls and defences? Of course not.

Despite any safeguards, always assume that everything you feed into AI will one day be freely available online. This might not be true, but it’s a good principle to encourage caution.

Download the AI Policy Template Now

10 Commandments for AI Use in the Workplace

Protect your organisation by following these rules for AI in the workplace.
You can also download our Free AI Policy Template, which incorporates all of the advice below in guidelines for your employees.
You may also find our AI training video useful.

1. Document and Audit Regularly

Maintain logs of which AI platforms are being used, by whom, and for what purposes. Conduct regular audits to ensure compliance with your AI policies and identify any shadow AI usage that may have crept in. Organisations can identify which employees are using AI tools by deploying Microsoft Defender and Cloud App Security, or by talking to your CSaaS account manager if you are equipped with Intersys’ Cyber Security as a Service.

2. Pay with Money, Not Data

Remember that free accounts will almost certainly be monetising your data and may not keep it private. Some are even designed to exfiltrate / steal information. Paid accounts are far more likely to respect your data – but always investigate terms (see next point).

3. Read the Fine Print

Different paid platforms offer varying degrees of privacy. For instance, ChatGPT’s paid Team plan does not currently train its models off your data, although some of its other plans do. Proceed in accordance with your risk tolerance.

4. Be an AI Control Freak

Only subscribe to AI platforms that allow you to centrally manage accounts. This means you can approve who should and shouldn’t access AI tools in your organisation.

5. Take Control of Settings

Some platforms give you a choice to use your data to ‘improve the model for other users’. Always toggle this off.

6. Don't Get Personal

Do not connect your organisation’s data with personal ChatGPT accounts - for instance, by providing the AI tool with access to Microsoft SharePoint apps.

7. Don’t Give Away the Farm

Redact all sensitive information before submitting to chatbots (find and replace is handy for this). Your employees should never submit your organisation's name or employee names to AI queries and always use placeholders such as ACME / Jane Doe instead.

8. Know Your Legal Boundaries

Understand which types of client data, proprietary information or regulated content (according to GDPR, HIPAA, etc) should never be submitted to AI platforms, regardless of their privacy promises. When in doubt, don't submit.

9. Fact Check Incessantly

AI can make mistakes, lie and display extreme bias. All AI outputs require thorough checking from a separate source (and not another AI). Missing this step could be extremely embarrassing… or worse.

10. No training, no Access

Ensure no employee has access to any AI platform until they have received appropriate training and signed your AI Policy.

What Do Thought Leaders About AI Say?

Matthew Geyman, MD of Intersys, contributes to many mainstream and industry publications on cyber security, AI and IT. Here’s his take:

‘In my opinion, if you’re implementing AI at pace, assigning a chief AI officer, or ensuring that your CTO is formally tasked with AI governance — and has the resources to respond – could and quite probably should be considered an urgent matter. Many businesses are already taking this on board. As reported in Insurance Times, LinkedIn reports that "since December 2022, 13% more firms have created head of AI positions".’

Matthew Geyman, Managing Director, Intersys

Download our Free AI Governance Policy Template

Help protect your organisation and show a readiness to comply with emerging legislation by using our Free AI Policy Template.
To receive a free copy of our AI governance policy template as well as our newsletter and updates, please enter your name and email in the box.

AI in the Workplace: Consultancy and Training

Intersys provides comprehensive AI cyber security training to both general users and IT departments. We can help you develop bespoke AI usage policies and assist your IT department in integrating tools safely. We also provide training to teams across organisations, to ensure safe and compliant AI use. Get in touch to find out more.

Talk to Intersys About AI Now

Managed IT Support

More information

Consulting Services

Executive IT Expertise

Compliance

Development

Risk

Cyber Security

Managed Security Services

Education Security

Further Services

IT Solutions

Installation

Microsoft Support

Troubleshooting

Telephony

Website Setup