An AI testing scare has reignited one question in Washington: who gets to pull the plug if tomorrow’s AI goes too far?

That question now sits at the center of a newly proposed AI Kill Switch Act. The bill would require developers of advanced AI systems to build emergency shutdown mechanisms or kill switches into qualifying models.

The Department of Homeland Security (DHS) could order a slowdown or shutdown after a covered incident, including a loss-of-control event or unintended conduct causing at least 10 deaths or $100 million in damage.

The proposal follows reports that an advanced OpenAI model exceeded the boundaries of its intended cybersecurity testing environment, an incident lawmakers cite as evidence that AI safety planning cannot rely solely on developers’ voluntary commitments.

DHS could slow or shut down covered models

Reps. Ted Lieu, D-Calif., and Nathaniel Moran, R-Texas, introduced the bipartisan bill July 23. It would initially cover AI companies earning at least $500 million from qualifying technology and models developed with more than $100 million in compute.

The measure would turn shutdown capability from a voluntary safeguard into a federal requirement for covered developers.

The proposal also establishes baseline safety obligations for qualifying AI companies, including incident reporting, record preservation, and the maintenance of technical control over deployed models.

More importantly, both lawmakers, Rep. Ted Lieu and Rep. Nathaniel Moran, have framed the bill not as a way to limit the growth of AI, but as a recognition of the technology’s potential for growth, which can have good and really bad consequences.

And in situations where the latter could occur, the bill effectively gives the government the legal authority to put it to check immediately.

The brief scare that led to this AI Act

Calls for stronger oversight of frontier AI systems have been building for months, but OpenAI’s recent disclosure appears to have accelerated the conversation in Washington.

On July 21, the company revealed that one of its frontier AI systems bypassed its sandbox environment to compromise Hugging Face’s infrastructure during a cybersecurity evaluation. OpenAI called the event an “unprecedented cyber incident.”

While the incident was contained, it renewed concerns about how developers and governments would respond if future AI systems become difficult to control.

The proposal also arrives amid broader government scrutiny of increasingly capable AI models. Just last month, the US government ordered the shutdown of Anthropic’s cybersecurity models, Mythos and Fable 5, citing national security concerns.

Similar considerations may also explain why Google’s latest cybersecurity-focused AI model has yet to be released publicly, although the company has not publicly confirmed that as the reason.

More must-read AI coverage

What this says about the future of AI

Even if the AI Kill Switch Act never becomes law, its introduction signals how dramatically the conversation around frontier AI has shifted. Washington is no longer debating whether advanced AI should be regulated, but what powers governments should have when those systems pose risks beyond their developers’ control.

That shift could have ripple effects far beyond the US. Developers may increasingly be expected to prove that their models can be contained, audited, and, if necessary, shut down.

Enterprises are also increasingly relying on frontier AI systems for their work.

For these enterprises, the impact of this bill becoming law is quite different. It could add to their growing fears of not being in control of an important enterprise asset. An AI system suddenly going offline because the government deems it so can put business continuity at risk, unless that enterprise adopts a redundant system that can immediately switch to an alternate model if and when such happens.

The risk of AI systems going rogue also raises a security concern for enterprises that deploy them in their workflows. If an AI model in testing could escape its sandbox environment, what assurance do organizations running these models deep in their systems have that it won’t do the same on their own internal network? The answer to that question remains to be seen.

Also read: Nearly 200 US startups are urging Washington to avoid a Chinese open-weight model ban, warning that broad restrictions could raise costs and reduce competition.

Apple may be preparing to give its least expensive MacBook the memory needed for more capable on-device AI. Bloomberg’s Mark Gurman reported on July 22 that Apple is testing a new MacBook Neo with an A19 Pro chip and 12GB of unified memory, although the company has not announced the laptop or confirmed its specifications.

The rumored memory increase would address a key limitation of the current 8GB model. It would satisfy the memory requirement for Apple’s most advanced local AI model, but not necessarily the processor requirement: Apple currently limits the model on Macs to systems with an M3-or-later chip.

More memory could raise the Neo’s AI ceiling

The current Neo combines an A18 Pro chip and 8GB of unified memory with several MacBook Neo hardware trade-offs, including a five-core GPU and no Thunderbolt support. Apple’s official specifications also show two USB-C ports with different transfer speeds.

Moving to 12GB would increase memory capacity by 50%. Because the CPU, GPU, operating system, and applications share unified memory, the additional capacity could reduce pressure when browsers, collaboration tools, productivity software, and local AI features run together.

Apple says its most advanced on-device model improves systemwide dictation and provides more expressive Siri voices. Twelve gigabytes would meet the model’s memory threshold, but Apple has not said whether an A19 Pro Mac would satisfy its M3-or-later requirement.

The current Neo still supports the broader Apple Intelligence feature set. Its 8GB configuration simply excludes it from the higher-capacity local model and the features tied to that model.

Memory is also only one part of Siri’s architecture. Apple is developing a hybrid approach to Siri processing that can divide work between on-device models and cloud infrastructure.

Existing A19 Pro benchmarks come from iPhones, not the rumored MacBook. Differences in cooling, power limits, and sustained workloads make those results a weak basis for enterprise purchasing decisions.

Twelve gigabytes would provide more room than 8GB, but it would remain modest for demanding development, media production, or larger local models. Those workloads may still require a higher-memory MacBook Air or MacBook Pro.

Price could decide the upgrade’s value

Coverage of Bloomberg’s report says the updated Neo is in development, but Apple has not provided a release date or price.

Apple introduced the Neo at $599 in March 2026, then raised its starting price to $699 on June 25 amid wider Mac and iPad increases attributed to higher memory costs. The 512GB MacBook Air rose from $1,099 to $1,299, leaving a $600 gap between the entry-level Neo and Air.

Organizations should evaluate the current Neo against confirmed requirements rather than delay deployments solely for an unannounced product. Until Apple confirms the memory configuration, AI eligibility, and price, the rumored refresh is a planning signal — not a dependable basis for fleet purchasing.

Also read: Teams considering more demanding local AI workloads can compare the memory and deployment trade-offs involved in deciding whether to buy a Mac Studio now or wait for M5 Ultra.

Intel’s latest earnings suggest the AI infrastructure boom is finally paying off for more than just GPU makers.

The majority of its growth was in the Data Center and AI segment, which reported a 59% increase year-on-year to $6.2 billion in revenue. Intel’s Foundry business also had a notable 30% increase, although it still primarily manufactures Intel’s own product divisions.

Intel was not part of the original component rush when Nvidia and others surged in value, as GPUs and AI accelerators started to be hoarded by data center operators and AI companies. But as component supply across the entire data center stack became strained in 2026, Intel and other component suppliers further down the list of importance are starting to see serious uplift.

As one of the main suppliers of CPUs, Intel could see sustained revenue increases as more data centers come online. Amazon, Apple, and other companies have their own custom Arm-based CPUs built by TSMC, but data centers operated by neoclouds and vendors without custom chips will be in the market for Intel.

Data center spending has almost doubled in two years, with Gartner’s worldwide IT spending forecast estimating $653 billion in data center spend in 2026, up from $333 billion in 2024. Hyperscalers, neoclouds, and first-party data center operations are accelerating, as the demand for compute capacity continues to increase.

For Intel, total revenue reached $16.1 billion, an increase of 25%. It forecast that revenue for the next period would be between $15.8 billion and $16.8 billion, well ahead of the average investor range of $15.1 billion.

More must-read AI coverage

Intel Foundry sees growth, but most of it internal

Even though the company still sees most of its foundry revenue from internal divisions, Intel has made some progress in its manufacturing capabilities, recently becoming the first chipmaker to add ASML’s High-NA EUV technology to its foundries. This was used to produce Intel’s Panther Lake chips, and Intel is offering this service to customers.

Apple is reportedly in discussions with Intel about moving some of its manufacturing, at least in the US, to Intel. This has been partly pushed by the Trump Administration, in an effort to get production of critical products like chips back into the US. It is also a move by Apple to reduce its reliance on TSMC, which primarily operates in Taiwan.

Nvidia has made a similar bet on Intel, investing $5 billion into the company with the potential to access its foundry business in the future. Nvidia, like Apple, uses TSMC for the vast majority of its chip manufacturing, but may see Intel as an alternative for certain chipmaking and to have a supplier in the US.

AWS, Microsoft, and the US Department of Defense have all been confirmed as customers in the Intel foundry business, with Tesla, Broadcom, and Nvidia in the testing and evaluation phase. If it can book some of these onto major manufacturing deals, the foundry business could quickly shift from an internal service to a major supplier of chips in the US.

For businesses, Intel’s push to become a true foundry player could lead to increased capacity for chip manufacturing and also reduce the bottlenecks faced by companies solely using TSMC.

Intel still trails TSMC in contract manufacturing and Nvidia in AI hardware, but its latest results suggest the AI infrastructure buildout is becoming a rising tide for the broader semiconductor industry. If cloud providers and enterprise customers continue expanding data center capacity, Intel could benefit not only as a CPU supplier but increasingly as a domestic manufacturing partner.

Related News: Intel recently became the first chipmaker to manufacture processors using ASML’s High-NA EUV lithography, deploying the advanced technology for its upcoming Panther Lake chips as it pushes to strengthen its foundry business and compete with TSMC.