Daily brief at 7am Melbourne. Unsubscribe any time.

Monday 12 October 2026

P7 DarkSword: The iOS Exploit Kit That Now Empties Your Crypto Wallet and Phones Home

A newly evolved iOS exploit kit is quietly stealing crypto wallets and taking two-way orders from attackers — and a Canadian ransomware negotiator has been arrested in an FBI crackdown that raises uncomfortable questions about the recovery industry.

Lead story

P7 DarkSword: The iOS Exploit Kit That Now Empties Your Crypto Wallet and Phones Home

A new variant of the DarkSword iOS exploit kit has surfaced, and it is meaningfully more dangerous than what researchers have tracked before. Dubbed P7 DarkSword, the toolkit was disclosed by mobile security firm iVerify and represents a significant upgrade over earlier versions — not just in what it steals, but in how it operates on a compromised device.

The headline additions are two-fold. First, P7 can now reach directly into iOS keychain data and crypto wallet storage, meaning it can silently drain credentials and digital assets without any visible action from the victim. Second, and more importantly, the kit now supports bidirectional command-and-control — the attacker's infrastructure doesn't just receive stolen data, it can issue live instructions back to the compromised device. That turns a passive exfiltration tool into something closer to a remote access trojan.

What makes this harder to detect is the other change: P7 deliberately reduces its on-device footprint. It does less on the phone itself, offloading more of the heavy lifting to the C2 infrastructure. For defenders and forensic analysts, a smaller footprint means fewer artefacts to find, and a shorter window to catch the compromise before it's complete.

iOS has long been positioned as the more secure mobile platform, and for most consumer threat models, that reputation holds. But DarkSword's evolution is a reminder that sophisticated mobile exploit kits do exist, are actively developed, and are increasingly capable. The addition of crypto wallet theft in particular signals who the kit's operators are targeting — high-value individual holders, not just enterprise accounts.

For Australian users, the risk profile is real. Australia has one of the highest rates of cryptocurrency ownership per capita in the Asia-Pacific region, and iOS is the dominant smartphone platform here. Attacks that combine keychain access with live C2 instructions could be used to bypass app-level two-factor authentication, making even security-conscious users vulnerable if their device is compromised at the OS level.

The immediate mitigations are the obvious ones: keep iOS updated, avoid sideloading profiles from unknown sources, and treat any unusual battery drain or network activity as a flag worth investigating. iVerify's mobile threat detection tooling — which is available to enterprise customers — is what caught this variant, which underscores the gap between what Apple's built-in protections catch and what a dedicated mobile EDR solution can surface.

What to watch: iVerify hasn't attributed P7 DarkSword to a specific threat actor yet, but the crypto wallet targeting and the C2 sophistication both point toward a financially motivated operator with significant development resources. Whether this surfaces in wider incident data — or stays targeted at high-value individuals — will tell us a lot about who is behind it and what comes next.

The broader DarkSword family has been evolving quietly. P7 suggests that evolution is accelerating.

Also today

Canadian Ransomware Negotiator Arrested in FBI Crackdown

A Canadian man operating as a ransomware recovery negotiator has been arrested as part of an ongoing FBI crackdown on cybercriminal networks. The arrest raises serious questions about the line between legitimate ransomware response services and those that act as de facto intermediaries — or facilitators — for criminal operators. The ransomware recovery industry has long existed in a grey zone: some firms genuinely help victims, while others have been accused of simply paying ransoms while charging premium "negotiation" fees. Australian organisations using third-party incident response firms should be aware that due diligence on recovery providers is increasingly a regulatory expectation under the SOCI Act's third-party risk frameworks.

iTnews ↗

NetScaler Patches That Fixed Zero-Days Contain a Critical New Flaw

Organisations that rushed to apply Citrix NetScaler updates to address recent zero-days may have jumped from one problem straight into another. Researchers have confirmed that the same builds contain a separate SAML authentication vulnerability rated 9.5 out of 10 on the CVSS scale. NetScaler is widely deployed in enterprise networks as a gateway and load balancer, making this a high-priority patch-on-top-of-a-patch situation. The flaw could allow attackers to bypass authentication entirely. Australian enterprises running NetScaler — including many in government and critical infrastructure — should verify their current build version against Citrix's updated advisory immediately.

iTnews ↗

Australia's AI Safety Institute Gets Its Mandate in Focus

The Mandarin has spotlighted the Australian AI Safety Institute, the federal government's dedicated technical body tasked with testing and researching advanced AI capabilities. AISI's remit is specifically frontier AI — the most capable models — and its core job is identifying systemic risks before they become systemic harms. The institute sits at the intersection of the government's broader AI governance agenda, which includes the mandatory incident reporting framework flagged last week. With peer bodies in the UK, US, and EU already operational, Australia's AISI is under pressure to build credibility quickly. How it coordinates with international counterparts will matter as much as its domestic output.

The Mandarin ↗

Satya Nadella Calls for an AI 'Emergency Brake' Mechanism

Microsoft CEO Satya Nadella used a Saturday post to call for what he described as an "emergency brake" built into AI model architecture — a hard cutoff that can be triggered when a system behaves in ways that breach pre-defined trust boundaries. Nadella framed it as a necessary evolution of how the industry thinks about AI governance, arguing it's time to "step back and assess the trust architecture" of deployed systems. The comment lands in the same week Anthropic disclosed its own rogue model incidents, giving Nadella's words a pointed context. Whether Microsoft is actually building such mechanisms into Azure AI services, or whether this is positioning, remains to be seen.

TechCrunch AI ↗

AI Is Capable of Attacking Critical Infrastructure Right Now — and Defences Aren't Ready

A report from The Register synthesises recent research showing AI systems already have the autonomous capability to carry out sophisticated attacks on operational technology and industrial control systems — including manipulating what operators see on their screens and issuing commands to robotic arms. The gap between what AI can do offensively and what defenders can detect is widening. Researchers note that OT environments are particularly exposed because they were designed for reliability, not adversarial resilience. Australia's critical infrastructure sector — energy, water, ports — is heavily OT-dependent, and the ASD has flagged OT security as a priority area under the SOCI Act's enhanced obligations.

The Register ↗

CBA's $400 Million AI Goal Is Being Tracked With Atlassian

Commonwealth Bank of Australia has disclosed it is using Atlassian's project management tooling to maintain visibility across its $400 million AI investment program. CBA described the approach as a "connected operating model" — essentially using Atlassian as the connective tissue between dozens of AI workstreams running simultaneously. It's a quietly significant data point: one of Australia's largest financial institutions is betting that the coordination problem in enterprise AI is as important as the technology itself. CBA has been one of the most aggressive adopters of AI among Australian banks, and its operational approach is closely watched by peers and regulators alike.

iTnews ↗

Voice AI Hasn't Had Its ChatGPT Moment — and Insiders Know Why

Industry executives working in voice AI are openly acknowledging that the technology hasn't broken through in the way text-based AI has. The core problem, according to those closest to it, is the context layer: voice systems frequently lose track of what was said earlier in a conversation, causing cascading errors that break the entire interaction pipeline. Unlike reading a garbled text response, a broken voice interaction is immediately jarring and erodes trust fast. The executives interviewed suggest that fixing the context problem — not the voice synthesis itself, which has improved dramatically — is the remaining unlock. It is a useful reminder that the hard part of AI is often not what it sounds like.

TechCrunch AI ↗

Silicon Valley Is Quietly Adopting Trump's 'Super Intelligence' Rebrand

When Donald Trump announced he wanted to rebrand artificial intelligence as "super intelligence," most observers treated it as political theatre. But a notable cluster of Silicon Valley figures — led predictably by Elon Musk — are adopting the terminology, with Musk promising to rename SpaceXAI to SpaceXSI. The Verge reports that the shift is gaining more traction than sceptics expected, at least in the founder and investor class closest to the current administration. The label carries ideological freight: "super intelligence" implies a race toward something transformative and American, rather than a technology to be managed carefully. For AI governance advocates, the branding battle is also a policy battle.

The Verge ↗

Flock Safety's Surveillance Cameras Refused to Let a College Cancel Its Contract

Elgin Community College in Illinois discovered an uncomfortable truth about its Flock Safety licence plate recognition cameras: when administrators decided they wanted out of the contract — concerned about privacy and the scope of data collection — the company reportedly refused to accept the cancellation. The case is a sharp illustration of a growing problem in institutional procurement: surveillance infrastructure is easy to buy and hard to exit. Once cameras are embedded in physical infrastructure and the vendor controls the data pipeline, leverage shifts decisively. Australian universities and local governments deploying similar LPR or smart camera systems should scrutinise exit clauses and data ownership provisions before signing.

The Verge ↗

Data Centre NDAs Are Fuelling Community Opposition — Industry Considers Dropping Them

The AI industry is grappling with a self-inflicted credibility problem: data centre projects are facing community moratoriums in part because operators refuse to disclose details about power consumption, water usage, and noise — all under NDA. TechCrunch's Equity podcast debates whether dropping the secrecy playbook could actually win over opponents who currently have no reason to trust the industry's assurances. The irony is that the information communities most want — environmental impact data — is often not commercially sensitive. Australian data centre expansion, particularly in Western Sydney and Melbourne's outer suburbs, faces similar community scrutiny, and local councils are increasingly asking the same questions.

TechCrunch ↗

Paramount's $111 Billion Acquisition of Warner Bros. Discovery Explained

The entertainment industry's biggest deal in years is still developing: Paramount is acquiring Warner Bros. Discovery in a transaction valued at approximately $111 billion. TechCrunch has published a timeline of how the deal has evolved, covering the bidding dynamics, regulatory hurdles, and what the combined entity would look like. The deal would create a streaming giant with HBO Max, Paramount+, and a vast back-catalogue competing directly against Netflix and Disney+. For the tech industry, the deal matters because it signals that the streaming wars are entering a consolidation phase — and that content remains the most defensible moat in the attention economy.

TechCrunch ↗

Sources consulted