Most AI news is noise. These are the signals worth acting on.
Risk, security, safety, regulation and implementation are moving at different speeds. We track the developments that change what organisations should control, approve, build, monitor or stop.
Updated
20 August 2026
For leaders accountable for AI risk, security, safety and governance.
Evidence first. Action second.
Current signal20 Aug 2026 · UK Legislation / DSIT
The UK just made AI governance harder to postpone.
The Data Protection Act 2018 (Code of Practice on Artificial Intelligence and Automated Decision-Making) Regulations 2026 require the ICO to produce a code of practice on good practice for the use of personal data in AI and automated decision-making. The instrument is current in August 2026.
What this means: the next layer of UK AI governance is moving from principle to operating guidance. Firms that wait for the final code will be assembling evidence retrospectively, for decisions already made. The cheaper path is to build the record while the systems are still being chosen.
01Inventory every AI system that touches personal data, including tools bought by individual teams.
02Define where human intervention is meaningful, not nominal, and name the person accountable.
03Log inputs, decisions, review points and escalation paths now, so the record exists before the code lands.
Regulatory Watch20 Aug 2026 · UK Legislation / DSIT
The UK just made AI governance harder to postpone.
The Data Protection Act 2018 (Code of Practice on Artificial Intelligence and Automated Decision-Making) Regulations 2026 require the ICO to produce a code of practice on good practice for the use of personal data in AI and automated decision-making. Alacrix view: the next layer of UK AI governance is moving from principle to operating guidance. If your AI touches personal data, the sensible move is not to wait for the final code. Build the evidence now.
What to do now
Inventory AI systems using personal data. Define where human intervention is meaningful. Log decisions, review points and escalation paths before the code tells you to.
Legal AI has a new regulatory fast lane. Evidence is still the entry ticket.
Applications opened on 5 August 2026 for the government's first Advisory AI Growth Lab, with legal services as the first sector, and run until 27 September 2026. The SRA is working with the Legal Services Board, the Council for Licensed Conveyancers and the ICO. Alacrix view: the signal is bigger than the programme. Regulators are moving from warning firms about AI to helping them test and deploy it responsibly, and the firms that benefit will arrive with a defined use case, clear controls and something worth testing.
Participation does not amount to regulatory endorsement or safe harbour.
What to do now
Pick one high-value legal workflow. Map the data, human review and evidence trail. Move it from experiment to governed pilot.
AI adoption is rising. Depth is not. That gap is your opportunity.
In the ONS release of 20 July 2026, covering businesses with 10 or more employees: around 35% reported use of at least one AI technology, 55% of employees reported using AI for work or education, only 10% of AI-using businesses reported using it extensively, and large language models were the most widely used technology at 18%. Alacrix view: when employee use runs ahead of enterprise adoption, you do not have an AI strategy, you have distributed experimentation.
What to do now
Find the three AI behaviours already happening inside your firm. Measure frequency, value and risk. Standardise the one with the best economics.
Mid-tier accountancy does not need more AI. It needs clearer reasons to use it.
ICAEW's 2026 mid-tier research ranks the leading enablers of increased AI adoption as relatable use cases (77%), better integration into existing tools (69%), training and upskilling (63%) and clarity on rules and regulations (54%). Alacrix view: buying another tool is not an adoption strategy. Adoption improves when AI becomes part of how the firm already works.
What to do now
Stop starting with vendors. Start with a recurring task that is slow, review-heavy or inconsistent. Then design the control and implementation around the workflow.
Your AI may now need to tell people it is AI. That is an interface problem, not a policy problem.
The Commission published guidelines on 20 July 2026. Article 50 transparency obligations apply from 2 August 2026 for certain providers and deployers of AI systems, including informing people when they are interacting directly with an AI system and transparency around certain AI-generated or manipulated content. Alacrix view: transparency obligations are becoming product requirements. If disclosure only exists in a policy PDF, it does not exist where the customer experiences the system.
Applicability depends on whether the firm and system fall within scope. Assess scope for the specific system and jurisdiction.
What to do now
Review every customer-facing assistant, generated communication and automated interaction. Put the disclosure at the point of use where required, not in a forgotten governance folder.
Agentic AI guidance is coming. Your controls should arrive first.
The ICO lists Agentic AI guidance as being drafted, with final guidance due Winter 2026. Alacrix view: an agent that can act is a different governance problem from a chatbot that can answer. The question is no longer only what the model can say, it is what the system is allowed to do without a person.
What to do now
Define authority limits, approved tools, data boundaries, human override, logging and incident rollback before autonomous workflows move into production.