Risk, Crisis and Emergency Communication Weekly

Warning-system testing, cross-border preparedness, AI governance and the evidence needed to turn risk information into action.

Coverage period: 08:00 on 20 July 2026 to 07:59 on 27 July 2026, Europe/Lisbon time. This episode reviews material first published or released in that period. It is a curated professional briefing, not a general news roundup.

This week in one paragraph

Across the selected material, the most useful shift is from message production to system performance. A warning system can meet a tower-coverage target and still leave people unable to hear, retrieve or act on the message. A review of disaster-risk research in the Greater Bay Area similarly links risk information to action only when people can understand it and access feasible protective choices. The operational reports from Venezuela and Yemen treat communication as part of response infrastructure: translated messages, community intermediaries, feedback routes and information-management roles sit alongside logistics and clinical work. The corporate breach and food-safety investigation show the other side of the problem: communicating what is known, what is not yet known and what people should do next while the factual picture is changing. Finally, the African AI-governance analysis is a reminder that human review and accountability need budgets, mandates and participation, not just principles.

Three developments to prioritise

1. AusAlert met an infrastructure benchmark, but not every user's information need

Sources: ABC News; The Guardian
Published: 27 July 2026
Type: Live reporting and explanatory analysis of a national warning-system test
Region: Australia
What was published / happened: ABC reported that the nationwide AusAlert test reached 94 per cent of targeted mobile towers, against a government benchmark of 90 per cent, while some people in regional areas or with older, imported or unupdated devices did not receive it. The Guardian recorded concerns from domestic-violence advocates about hidden or "safe" phones and from accessibility advocates about the difficulty of taking in a short alert and returning to it later. The government said SMS, radio and television would remain part of the warning mix.
Why it matters for communicators: Infrastructure success is not the same as usable warning performance. The relevant test is end-to-end: receipt, audibility, comprehension, recall, accessibility, safe response and fallback when networks or devices fail.
Possible application: Add a warning-system exercise inject in which the message arrives on an old device, during a network outage, or to someone who cannot safely expose a phone. Test the fallback script and the time needed to repeat the action advice.
Evidence note: The tower figure is an official result reported by ABC; device and accessibility concerns are reported statements and user experience, not an independent evaluation.

2. Jordan began a multilingual operational trial before results were available

Source: Petra, the Jordan News Agency
Published: 27 July 2026
Type: Official news report on an operational trial
Region: Jordan, Middle East
What was published / happened: Jordan's National Center for Security and Crisis Management began trials of its national mobile warning-message system across most governorates from 27 July to 2 August. The stated tests covered delivery speed, message effectiveness and overall early-warning efficiency. Messages were to be sent in Arabic and English with an audible alert, and the public was told not to panic when a test message arrived.
Why it matters for communicators: The useful part is the test design, not a claim of success. Delivery speed and network reach are necessary but do not tell a crisis team whether people understood the language, recognised a test, or knew what action would follow a real alert.
Possible application: Require every alert trial to capture comprehension and action-choice measures by language and governorate, then publish the limitations as well as the coverage result.
Evidence note: Official account of an in-progress trial; no outcome data were available in the window.

3. The first days of the Origin breach show why a dated knowledge timeline matters

Source: Origin Energy, Update July 2026
Published: 22 July 2026 for the first notice, with dated updates on 23 and 24 July
Type: Primary incident update and customer-support notice
Region: Australia
What was published / happened: Origin's first notice described a potential security incident and possible unauthorised access to customer data, while saying it did not then believe credit-card or bank-account information was involved. On 23 July it confirmed unauthorised access and disclosure of some data, said it was still identifying affected customers, and apologised. The 24 July update kept support open over the weekend. The page has since been updated, so this item uses only the dated entries inside the coverage window.
Why it matters for communicators: Early cyber-crisis language has to preserve accuracy without sounding evasive. A dated record should show the change from suspicion to confirmation, explain what remains unknown, warn about scams and give affected people a route to help.
Possible application: Rehearse a three-column incident log: confirmed facts, working hypotheses and next verification step. Link each customer-facing sentence to one of those columns and assign an owner for the next update.
Evidence note: The 22–24 July statements are primary but early and provisional; later page content is outside this episode's window and is not used here.

Research and evidence

4. A Greater Bay Area review links warnings to feasible protective choices

Source: Geng, Zhao, Yang and Wang, "Residents’ Responses to Urban Disaster Risks in the Guangdong–Hong Kong–Macao Greater Bay Area: A Systematic Review Using Bibliometric and Topic-Modelling Approaches", Urban Science
Published: 25 July 2026
Type: Open-access systematic review
Region: Guangdong–Hong Kong–Macao Greater Bay Area, China
What was published / happened: The authors screened 251 records and included 144 in a PRISMA-aligned review using bibliometric analysis, keyword co-occurrence, a region-hazard matrix and BERTopic-inspired modelling. The review identifies themes including risk perception and preparedness, evacuation and shelter accessibility, flooding, typhoon and coastal risk, community resilience, and health-emergency response. Its framework links risk information, cognitive appraisal, response action and community resilience, and argues that warnings need to be timely, intelligible, credible and accessible and connected to feasible choices.
Why it matters for communicators: The review does not show that any single message caused behaviour. It does show why warning design should be evaluated as a chain from information to decision, including cross-boundary terminology, transport information and digital vulnerability.
Possible application: Red-team a warning by asking: what protective choice does it make feasible, who cannot take that choice, and where will the resident find the next operational update?
Evidence note: Systematic review and bibliometric synthesis; the Greater Bay Area is an analytically informative case, not a statistically representative population study.

5. A conceptual risk-science preprint argues for participation under uncertainty

Source: McGranaghan and colleagues, "From risk to flourishing: Seeds for responding to a complex world", arXiv
Published: 22 July 2026
Type: Preprint, conceptual risk-science paper
Region: International / conceptual
What was published / happened: The authors revisit tensions between physical and social risk, determinism and uncertainty, and contextuality and generality. They propose a dialectical approach grounded in interdependence, responsivity, participation, pluralism, openness and ongoingness. It is not an evaluation of a crisis campaign, but it offers a useful challenge to communication teams that treat uncertainty as a defect to be hidden rather than a condition to be managed with affected communities.
Why it matters for communicators: The paper supports a shift from one-way certainty claims towards structured participation: explain what is known, what is contested, what could change and how communities can shape the next decision.
Possible application: Put a contested-risk scenario into a tabletop and require the team to publish a decision log with uncertainty, dissenting evidence, review triggers and named community input.
Evidence note: Preliminary conceptual preprint; no empirical sample or causal test.

6. A new MENA health-communication editorial maps a field, but is not new outcome evidence

Source: "Health communication in the Middle East and North Africa: review of current research and future directions", Discover Health Systems
Published: 23 July 2026
Type: Open-access editorial
Region: Middle East and North Africa
What was published / happened: This editorial reviews the direction of health-communication research in the region and points towards the practical importance of accurate information, crisis and risk communication, and locally grounded research. It is useful as a regional signal: health communication cannot be separated from trust, language, access and the information environment in which health decisions are made.
Why it matters for communicators: A regional research agenda is not a substitute for outcome evidence. Its value is in prompting teams to examine whose evidence base they use, whether imported models fit local media and health systems, and where local-language monitoring is missing.
Possible application: During an exercise, require a source-and-language map for each affected group, including trusted intermediaries and the evidence needed to adapt advice.
Evidence note: Editorial synthesis, not a new primary study or systematic review; the article page is open access but provides limited empirical detail.

Cases and current developments

7. The FDA closed a Salmonella investigation while keeping the action message concrete

Source: U.S. Food and Drug Administration, "Outbreak Investigation of Salmonella: Moringa Leaf Powder (January 2026)"
Published: 22 July 2026 update
Type: Official outbreak investigation update
Region: United States and territories
What was published / happened: The FDA said its investigation was complete and that the CDC had declared the outbreak over. The page recorded 131 illnesses, 36 hospitalisations and no deaths across 38 states, with 12 samples positive for Salmonella and 11 showing at least one outbreak strain on whole-genome sequencing. Consumers and retailers were told not to eat, sell or serve recalled products, to discard them, clean affected surfaces and seek care if symptoms developed.
Why it matters for communicators: Closure language did not erase the need for action. A final investigation update still had to connect evidence, product identification, recall behaviour and health advice, while making clear that the illness dates and the communication update were not the same thing.
Possible application: Build a recall template that has separate fields for investigation status, product scope, action verbs, symptoms, refund route and what remains under surveillance.
Evidence note: Official investigation update, supported by laboratory and genomic testing cited on the page; the update is not an independent evaluation of public comprehension.

8. Venezuela's earthquake response treated risk communication as an operational function

Source: PAHO/WHO, "Situation Report No.7: Earthquakes in Venezuela 2026 (M7.2 and M7.5), 23 July 2026", via ReliefWeb
Published: 24 July 2026
Type: Official situation report
Region: Venezuela, Latin America and the Caribbean
What was published / happened: More than three weeks after the 24 June earthquakes, PAHO/WHO described a shift from acute rescue and trauma care towards continuity of health services, temporary camps and early recovery. Its response team included health-emergency coordination, information management, risk communication and community engagement. The report also described a real-time surveillance dashboard, virtual emergency coordination and 14 emergency medical teams operating across affected areas.
Why it matters for communicators: Communication is strongest when it is connected to operational decisions: what services are open, what risks are being monitored, how referrals work and how communities can ask questions. A dashboard is useful only if someone turns it into understandable, local action information.
Possible application: Define the handover point between emergency and recovery communications, and test whether the same team can update service information, surveillance messages and community feedback without creating competing channels.
Evidence note: Official operational report; figures and activities are reported by the response system and were not independently evaluated in the document.

9. Yemen's early-action report makes translation and feedback part of warning readiness

Source: IFRC, "Yemen floods - Simplified Early Action Protocol - Final Report, May 2026", via ReliefWeb
Published: 24 July 2026 page publication; report period and title refer to May 2026
Type: Final report and lessons-learned document
Region: Yemen, Middle East
What was published / happened: The report describes a simulation exercise conducted with the IFRC, Yemen Red Crescent Society and Civil Defense before activation. It records translated and locally adapted early-warning and mitigation messages, posters and brochures, community meetings, megaphones, door-to-door outreach, and complaints and feedback mechanisms. A post-distribution survey covered 249 households, representing 1,876 people.
Why it matters for communicators: The report treats message adaptation, community intermediaries and feedback as readiness work, not as an afterthought once the alert has been sent. It also shows why the exercise should test the route from warning to trusted local action.
Possible application: In the next flood tabletop, assign a community-feedback lead and require the team to change one message in response to a plausible misunderstanding, not merely to deliver a pre-written script.
Evidence note: Implementer-reported final report with retrospective lessons; the sample is a monitoring exercise, not a controlled communication-effect study.

10. The Western Balkans report shows that early warning also needs a governance map

Source: UNDRR, "Resilient Western Balkans: Strengthening transboundary collaboration and capacities on disaster risk reduction and Early Warning Systems in the Western Balkans — Year 1 Narrative Report"
Published: 24 July 2026
Type: Year-one narrative report and evaluation document
Region: Albania, North Macedonia, Serbia and the wider Western Balkans
What was published / happened: UNDRR's report describes a regional initiative to strengthen transboundary collaboration and early-warning capacity. It records a steering committee in Serbia, a project working group in North Macedonia and focal points within Albania's civil-protection agency, while noting that the formal governance arrangement in Albania was still being discussed.
Why it matters for communicators: Multi-agency communication fails quietly when ownership is unclear. A warning may be technically interoperable but still lack a named sign-off authority, a shared vocabulary or an agreed route for transport, border and public-information updates.
Possible application: Add a governance test to the next exercise: who can approve the first message, who can translate it, who can correct it, who briefs elected leaders and who owns the public explanation when agencies disagree?
Evidence note: Year-one implementation report; it records structures and progress, not evidence that the arrangements have yet improved public outcomes.

Policy, regulation and AI governance

11. An African AI-governance analysis puts implementation before principle statements

Source: Paul Kimumwe, "Rethinking Africa’s Approach to the Politics of AI Governance and Regulation", CIPESA
Published: 23 July 2026
Type: Identifiable expert analysis
Region: Africa, with examples from Kenya, Rwanda and Ethiopia
What was published / happened: Kimumwe argues that African AI policies and laws can stall when political commitment, public participation, funding and regulatory coordination are weak. The analysis contrasts 40 African countries with data-protection laws against 16 that had ratified the Malabo Convention, and notes that Kenya's 2025–2030 AI strategy estimated implementation costs of 152 billion Kenyan shillings without a dedicated allocation in the 2026–27 budget review discussed. It also describes fragmented responsibilities across agencies in several countries.
Why it matters for communicators: AI-assisted crisis communication needs more than a human-in-the-loop sentence in a policy. Teams need a funded use-case register, clear approval rights, data-protection controls, local participation and a route to suspend an automated output when the situation changes.
Possible application: For every AI-supported warning or monitoring use case, record the accountable human owner, data source, affected languages and groups, review threshold, audit trail, fallback channel and budget line.
Evidence note: Expert analysis drawing on cited policy and budget material; the claims are not an independent comparative evaluation.

What this means for communicators

Test the public task, not only the delivery mechanism. Measure receipt, comprehension, recall, accessibility and action, including people on old devices, people with disabilities and people who cannot safely use a phone.
Make uncertainty legible. Maintain a dated knowledge timeline with confirmed facts, working hypotheses, unresolved questions and the next verification step. This applies to cyber incidents, outbreaks and conflict-related information alike.
Treat language and feedback as system components. Arabic and English alert trials, translated Yemen messages and the MENA research agenda all point to the same planning question: who can adapt advice locally and how will the team know that it was understood?
Put communications inside the response architecture. Venezuela's situation report names information management, risk communication and community engagement alongside clinical and logistical functions. Put those roles into the EOC chart and handover criteria.
Govern AI with resources and stop conditions. Define the human owner, audit trail, data boundaries and fallback before the first crisis prompt is used. If the system cannot explain where a claim came from, it should not be the source of an alert.

Watchlist

Australia's AusAlert feedback and the planned move from testing towards formal operation, especially accessibility and fallback-channel findings.
Jordan's alert trial results after 2 August, including delivery performance by language and governorate and whether the public understood the test message.
Further Origin Energy updates, as the affected-customer picture and support obligations develop beyond this episode's window.
Whether the warning and governance arrangements described in Venezuela, Yemen and the Western Balkans produce published after-action evidence rather than only implementation outputs.

Editorial log and transparency

Candidates reviewed included newly dated research, preprints, official investigations and situation reports, warning-system tests, AI-governance analysis and preparedness documents. We rejected older or newly updated work, a crisis-hotline paper published on 3 July, accepted-but-not-yet-published studies, an undated wartime-principals article, press-release-style notices, vendor material, event promotion and claims that could not be independently dated or evidenced. The 11 selected items were checked against the 20–27 July Europe/Lisbon window, clustered by underlying development and retained only where there was a concrete communication implication. Significant limits remain: several official documents are implementer-reported, and two selected editorials or analyses are not empirical outcome studies.

This episode was researched, selected and scripted with AI assistance. A human editorial review set the scope, checked dates and source pages, removed unsupported or out-of-window candidates, and made the final selection. The narration is synthetic. AI assistance does not turn a source claim into an independently verified fact; the evidence status and limitations above are part of the editorial record.

Sources

Qixiang Geng, Shufang Zhao, Hang Yang and Xi Wang. "Residents’ Responses to Urban Disaster Risks in the Guangdong–Hong Kong–Macao Greater Bay Area: A Systematic Review Using Bibliometric and Topic-Modelling Approaches." Urban Science, 25 July 2026. Type: Open-access systematic review. Region: China, Greater Bay Area. Access: Open. Direct source
Ryan M. McGranaghan, Elisabeth H. Krueger, Alexandre Dunant, Guillaume Falmagne, Lynee L. Turek-Hankins, Sara M. Constantino, Raissa D’Souza, Elke U. Weber and Graydon W. Manzke. "From risk to flourishing: Seeds for responding to a complex world." arXiv, 22 July 2026. Type: Preprint. Region: International. Access: Open. Direct source
Discover Health Systems editorial. "Health communication in the Middle East and North Africa: review of current research and future directions." Discover Health Systems, 23 July 2026. Type: Open-access editorial. Region: Middle East and North Africa. Access: Open. Direct source
ABC News. "AusAlert test: Millions of devices sent nationwide siren — as it happened." ABC News, 27 July 2026. Type: Live reporting. Region: Australia. Access: Open. Direct source
The Guardian. "AusAlert test will blast emergency warning siren over most phones in Australia on Monday. Here’s what to know." The Guardian, 27 July 2026. Type: Explanatory reporting. Region: Australia. Access: Open. Direct source
The Jordan News Agency, Petra. "Authorities Test National Mobile Emergency Alert System Across Most Governorates." Petra, 27 July 2026. Type: Official news report on an operational trial. Region: Jordan. Access: Open. Direct source
Origin Energy Limited. "Update July 2026." Origin Energy, first notice 22 July 2026, dated updates 23–24 July 2026. Type: Primary incident update and customer-support notice. Region: Australia. Access: Open; page now contains later dated updates, which were not used. Direct source
U.S. Food and Drug Administration. "Outbreak Investigation of Salmonella: Moringa Leaf Powder (January 2026)." FDA, 22 July 2026 update. Type: Official outbreak investigation update. Region: United States and territories. Access: Open. Direct source
Pan American Health Organization / World Health Organization. "Situation Report No.7: Earthquakes in Venezuela 2026 (M7.2 and M7.5), 23 July 2026." PAHO/WHO, via ReliefWeb, 24 July 2026. Type: Official situation report. Region: Venezuela. Access: Open. Direct source
International Federation of Red Cross and Red Crescent Societies. "Yemen floods - Simplified Early Action Protocol - Final Report, May 2026." IFRC, via ReliefWeb, 24 July 2026 page publication. Type: Final report and lessons-learned document. Region: Yemen. Access: Open. Direct source
United Nations Office for Disaster Risk Reduction. "Resilient Western Balkans: Strengthening transboundary collaboration and capacities on disaster risk reduction and Early Warning Systems in the Western Balkans — Year 1 Narrative Report." UNDRR, 24 July 2026. Type: Year-one narrative report and evaluation document. Region: Western Balkans. Access: Open. Direct source
Paul Kimumwe. "Rethinking Africa’s Approach to the Politics of AI Governance and Regulation." CIPESA, 23 July 2026. Type: Identifiable expert analysis. Region: Africa. Access: Open. Direct source

Risk, Crisis and Emergency Communication Weekly — 27 July 2026