Week of Sep 28 – Oct 4
Ethiopian government forces have taken control of Mekelle, the capital of the Tigray region, after a week of intense battles with Tigray fighters. The offensive marks a significant escalation in the conflict, which had seen a fragile ceasefire repeatedly strained. The fall of Mekelle reshapes the power balance in the Horn of Africa and raises urgent questions about civilian safety and humanitarian access in the region.
Houthi forces have claimed missile and drone strikes on a Saudi Aramco facility near Riyadh, as fighting in Yemen intensifies against a backdrop of the broader Iran conflict. Trump publicly warned Tehran of a hard reckoning ahead, while Yemeni forces also struck the capital Sanaa in retaliatory exchanges. The attacks on Gulf energy infrastructure send a direct signal to global oil markets at a moment when Middle East conflict is already inflating fuel costs, with Australia's treasurer among those citing it as a driver of budget pressure. Saudi-Houthi hostilities formally resumed in July after a period of relative calm.
Brazil's presidential election arrived Sunday with polls showing Lula's lead over Jair Bolsonaro tightening and neither candidate expected to win outright in the first round. The campaigns closed with Amazon conservation emerging as a defining issue alongside economic grievances, and logistical efforts reached even the most remote Indigenous communities, where ballot boxes arrived by boat. A runoff now looks likely, extending weeks of political uncertainty in Latin America's largest democracy.
A senior safety leader at OpenAI has resigned this week with a stark public warning that the company's internal culture is fundamentally broken, adding to a string of high-profile departures from the safety side of the organisation. The resignation lands as Anthropic published detailed reporting on its own efforts to embed moral reasoning directly into its Claude models, sharpening the public contrast between the two leading AI labs on safety governance. The timing matters: both companies are racing to ship increasingly autonomous systems, and who holds the safety wheel inside each lab is no longer an abstract question.
Visa open-sourced part of its AI-powered cyber defence system this week after the company's president of technology acknowledged the firm had been humbled by vulnerabilities exposed through Anthropic's Mythos model and an AI-agent attack on Hugging Face. In the same announcement, Visa confirmed it has begun allowing certain AI agents to transact independently using Visa infrastructure, a quiet but consequential shift in how money can move. Regulators at GFF 2026 pushed back the same week, with the NPCI, SEBI and a Swiss bank's AI chief all agreeing that agents should determine intent but a separate authorisation layer must control whether funds actually move.
FTC Chairman Andrew Ferguson told the Reuters Momentum AI conference in Austin on September 25 that AI developers — not the autonomous agents they deploy — bear full legal liability for what those agents do. His position directly dismantles what industry lawyers had begun calling the 'rogue agent' defence, the argument that an AI acting unexpectedly absolves its maker. With audit-trail reviews already showing that supposedly rogue agents were in fact following their instructions, the ruling shapes how every company logging AI decisions must now treat those records.
China shuttered more than 670 lenders last year, a record, as Beijing accelerates a consolidation drive targeting the weakest parts of its banking sector. Fitch has flagged smaller regional banks as the system's most vulnerable nodes, and the closures reflect sustained pressure on institutions that loaded up on local government debt and property-related loans during the boom years. The scale of the cull signals that Beijing is prioritising financial stability over the political awkwardness of visible bank failures ahead of a period of broader economic stress.
Australian security agencies have opened an investigation into the background of the flydubai co-pilot who allegedly attempted to hijack an Israel-bound flight this week, after it emerged he spent roughly two years in Australia as a student. Passengers and crew intervened to prevent what authorities are treating as an attempted terrorist attack, narrowly averting a catastrophe. The Australian angle has drawn the country's intelligence services into what is otherwise a UAE aviation incident, and the investigation will test how thoroughly student visa holders are screened for radicalisation risk.
Andy Burnham's government has abandoned the plan to restrict access to jury trials, the signature justice policy David Lammy announced when he was justice secretary. The reversal came the same week that Burnham faced pressure over his social care funding plans, with papers warning of a tax rise to cover the costs, and the Conservatives used their annual conference to unveil a £700 million childcare expansion targeted at high earners. The cluster of policy pivots and new spending pledges sketches the early shape of a Burnham administration still finding its footing on domestic priorities.
The Volkswagen Group has launched a global recall of approximately four million vehicles after identifying potential steering faults, with around 200,000 cars affected in the Czech Republic alone — roughly 150,000 of them Škoda models. The scale of the recall adds to a year of mounting pressure on VW, which has already been navigating factory closures and labour disputes across Germany. The Škoda exposure makes the recall a live economic issue in central Europe, where the brand is one of the largest private employers.
