Author: Anthony John Padilla

Bill Gates has sharpened his call for AI rules in 2026. He says the technology has moved past a few key thresholds, and the consequences are already showing up in fraud, cybercrime, hiring, and healthcare. He points to hacking, bioterrorism, and large-scale fraud, argues that some jobs should remain in human hands, and says the fight now is over who takes the risks, who gets the upside, and whether regulation shows up before the damage does.

Microsoft Copilot Download

He also says the dangers aren’t hypothetical anymore. Gates keeps coming back to hacking, bioterrorism, and fraud at scale, and the numbers around deepfake scams help explain why: attempts in 2026 were up more than 2,000% compared with 2022.

The policy picture he’s describing is uneven. The European Union’s AI Act brought in transparency rules on August 2, 2026. The US still had a patchwork approach. China, for its part, added ethics, registration , and security requirements in July 2026.

The labor data is mixed, but not hard to read. Goldman Sachs found slower growth in job openings across AI-exposed sectors in the US, Germany, and Australia starting in late 2022. The World Economic Forum, meanwhile, projected 85 million displaced jobs by 2026.

PwC found gains too: better productivity, higher wages, and in some cases more hiring. Still, new graduates are getting pinched as AI absorbs the kind of entry-level work people used to learn on.

If you follow AI policy, this is the part to watch:

Gates says some roles, diagnosis among them, should stay human even as AI spreads through medicine. By February 2026, the US Food and Drug Administration had cleared 1,357 AI medical devices and, as Gates notes, 81% of US physicians were using them, mostly for administrative work.

He’s also arguing that a few pieces of the debate still don’t get enough attention: developing countries, predictive policing, credit scoring, and the uneven effects on different groups. In his view, the argument has shifted. It’s now about who carries the risks, who captures the gains, and whether the rules arrive in time.

You can find Gates’s essays and interviews online.

In 2025, the Trump administration ramped up wild horse sales, with the Bureau of Land Management moving about 3,700 animals out of federal custody, more than twice the earlier level, according to BLM budget and sales records.

Current law bars the government from directly selling wild horses for slaughter. Still, when adoptions fall through, the BLM can sell them for as little as $25 under its sale policy, and advocates say hundreds later ended up resold to slaughterhouses in Canada and Mexico. That jump in sales sits inside a much bigger removal push. The BLM had about 20,000 gathers planned for FY2024, according to its budget documents, after removing roughly 50,000 wild horses and burros between 2020 and 2023, according to roundup data.

Then the budget proposals for FY2026 and FY2027 pushed the issue further, seeking to strip out language that blocks the killing of healthy horses and their sale for slaughter, according to BLM budget proposals. The strain shows up pretty clearly in the numbers:

As of March 2024, 73,520 wild horses and burros were still on public land, according to BLM population estimates. That was down from more than 82,000 in 2023, but it was still nearly three times the 27,000 target. Holding costs reached $108.5 million in 2023, according to BLM holding-cost records, and the number of animals in captivity was projected to pass 80,000 in 2024, according to BLM budget documents.

If you follow public-lands policy or animal-welfare fights, keep an eye on this. Welfare groups want more fertility control and less reliance on roundups and low-cost sales. Ranchers, along with some environmental groups, support stronger removals because of grazing pressure and ecosystem damage.

The figures are in BLM budget, population, roundup, and holding-cost documents.

Kaspersky says it found Android malware aimed at DoFun in-vehicle infotainment systems. The campaign hits Android-based DoFun head units and similar dash systems, pulling ad fraud and proxy-botnet activity into the dashboard itself.

Per Kaspersky, the malicious code shipped alongside legitimate DoFun software, which points to a supply-chain or packaging compromise that most users, installers, and resellers wouldn’t catch. It runs without any visible interface and quietly collects the device model, screen resolution, and Wi-Fi network ID. Not passwords or banking details, but enough to profile infected units, manage them, and make money from them.

It also matches the broader direction automotive security has been heading. Public automotive cyber incidents reached 494 in 2025. Ransomware made up 44% of them. Back in 2022, 97% of logged attacks were remote. VicOne says more than 77% of automotive vulnerabilities are in onboard systems, and IVI security spending is projected to climb from $2.4 billion in 2025 to $7.56 billion by 2034.

If you use a DoFun head unit, or depend on third-party or unofficial updates, don’t brush this off.

A growing group of scholars is warning that, in 2026, all the talk about “conscious” AI can serve a pretty convenient purpose for chatbot companies: it can blur responsibility when users get hurt. Their point is simple. It doesn’t matter much whether a company pitches a chatbot as a mere tool or something more like an independent actor. The question is who answers for the damage. And in their view, there isn’t some special accountability gap here. Negligence law and product-liability law can still reach the companies that build, market, and roll out these systems.

ChatGPT Download

The number of incidents keeps moving the wrong way. One database recorded 362 cases in 2025, up from 233 in 2024. An OECD tracker shows something even starker, a tenfold increase in incidents from early 2020 to January 2026.

Courts are starting to push on the issue too. Wrongful death and personal injury lawsuits are already testing claims tied to suicide and worsening mental health. In one case, a court has signaled that AI output could count as a product, which could chip away at Section 230 of the Communications Decency Act as a defense.

If you pay attention to AI policy, keep an eye on this.

Europe is moving toward risk-based rules that should apply by June 2026, along with a liability proposal. The US is taking a different route, leaning more on voluntary commitments and existing regulators. Inside companies, the picture is mixed. The share cutting AI use without a policy fell from 24% to 11%, but the problems haven’t gone away: 59% still point to knowledge gaps, 48% cite budget limits, and IBM says AI-related breaches cost an average of $6 million. The fight ahead is pretty easy to see. Who pays, who fixes the damage, and who should’ve stopped it before it happened.

Apple has shipped macOS Tahoe 26.6.2, along with iOS and iPadOS 26.6.1. The updates bring a sizable set of security patches for Macs, iPhones, and iPads, and most of them touch WebKit, the browser engine behind Safari.

For supported current devices, Apple lists 28 fixes. On the Mac side, 21 of those are tied to WebKit. iPhone and iPad users get those same 28 fixes, plus one more for Telephony authentication. Since every browser on iOS and iPadOS has to use WebKit, the bugs don’t just matter for Safari. Third-party browsers are affected too. The risks range from crashes and memory corruption to data leaks, code execution, sandbox escape, and data exfiltration.

Apple also pushed iOS 18.7.10 and iPadOS 18.7.10 for older devices. Those updates patch more than 120 bugs in total, based on Apple’s security notes, including more than 40 in WebKit. The rest cover areas like the Kernel, ImageIO, Audio, and networking, with issues tied to denial-of-service, private data exposure, code execution, and traffic interception.

Apple says it hasn’t seen any of these flaws used in active attacks. Still, once the security notes are out, unpatched devices get a lot easier to target. If your Mac, iPhone, or iPad can install one of these updates, it’s a good idea to do it.

You can download the updates now from Software Update on your Mac, iPhone, or iPad.

The White House has released a new National Security Presidential Memorandum that opens a formal path for vetted private cybersecurity firms to help go after foreign cybercrime. Any company brought in would operate under Department of Homeland Security and Department of Justice oversight.

With that approval, companies could be authorized on a case-by-case basis to conduct cyber surveillance and cyber effects operations against overseas ransomware gangs, phishing networks, and fraud operations. The focus is especially on foreign safe havens where arrests, extraditions, or direct law enforcement action can be difficult.

That’s a real shift from the U.S. government’s long-standing reluctance to let private actors do much beyond defense and incident response.

If you follow U.S. cyber policy, keep an eye on this one. Washington says cybercrime caused more than $4 billion in U.S. losses in 2020. Many of the strongest defenders sit outside government, most of the infrastructure being targeted is privately run,

and this move lines up with a broader 2026 strategy alongside anti-fraud and National Security Systems efforts, while adding to sanctions and cooperation aimed at threats Washington has tied to China, Russia, Iran, and North Korea.

You should also expect intense scrutiny around legality, ethics, oversight, liability, hack-back limits, cross-border mistakes, retaliation,

and the hard-to-predict consequences of letting private firms move from defense into disruption.

You can read the memorandum through official White House channels.

Libpng, the PNG image library sitting under a huge amount of software, got an update in February 2026 to fix CVE-2026-25646. This one goes all the way back to 1995. It was rated high severity, and it was also one of 17 vulnerabilities that stayed buried for more than a decade.

The flaw is a heap buffer overflow in the rarely used `png_set_quantize` function. A maliciously crafted PNG could crash an app and, in the worst case, expose information or open the door to remote code execution. Vulnerable libpng builds showed up in Debian, Red Hat, Ubuntu, desktop applications, and some Java runtimes. It’s a good example of how code written in the early Unix and DOS years still ends up inside current systems, even when actually exploiting the bug may not be easy.

PrintDemon follows the same script. It was a Windows printer weakness introduced in 1996 and not fixed until May 2020. What started as a permissive design choice to make printer setup easier turned into a security problem years later.

If you’re responsible for Linux, Windows, operational technology, or critical infrastructure, both the libpng fixes and the Windows patches are worth installing. Organizations still take more than 100 days on average to patch. Nearly 60% of compromises involve unpatched vulnerabilities. And older unsupported systems often don’t have strong access controls or modern encryption in place.

Get the patched versions from your Linux distribution, the affected application or Java runtime update, and Windows Update.