Lesen

Israeli security startup Rig Security launched on September 29 with $12 million in seed funding.

The blind spot it targets is specific: coding assistants and autonomous agents rarely hold accounts of their own, so their actions are logged under the employee or service account that launched them. In the words of CEO Guy Kozliner, agents are now the most active identities in an enterprise, "and they are acting under human names" — security teams cannot tell person from agent, and stopping the agent means blocking the employee too.

The product uses machine learning to match one identity across identity systems, with the company itself claiming better than 96% accuracy; an endpoint sensor can stop an agent's risky action without interrupting the employee. The company says Fortune 200 customers in financial services, insurance and health care already run it in production — both that claim and the accuracy figure are the company's own. The round was co-led by Ten Eleven Ventures and Brightmind Partners, with CrowdStrike participating as a strategic investor.

AbbVie, argenx, Lundbeck and Takeda have joined Berlin-based Apheris and Ginkgo Datapoints in a new Antibody Developability Consortium, training AI on roughly 10,000 antibody sequences to predict which drug candidates can be manufactured and reach the clinic.

Each member contributes proprietary sequences and trains models inside its own environment via Apheris's federated infrastructure, without raw sequences being exposed to other members. Ginkgo Datapoints fills remaining capacity from public sources and runs wet-lab characterisation; the initial dataset is due to members by early 2027.

The consortium has just kicked off and has released no model performance data. Charlotte Deane of Oxford and Peter Tessier of the University of Michigan provide independent scientific oversight.

The US Navy formally established the Robotic and Autonomous Systems Warfighting Development Centre in Virginia on September 24, tasked with turning its aerial, surface and undersea drones into an integrated fighting force.

The centre will develop doctrine and tactics for unmanned systems, train sailors, and test whether the platforms keep operating under realistic conditions. Chief of Naval Operations Admiral Daryl Caudle said years of robotic-systems work had proceeded separately across domains, that the next step is integration, and that "a prototype is not combat power."

US commanders have previously linked this drone capability to disrupting a PLA attack on Taiwan. The centre's budget and staffing have not been disclosed; its integration progress remains to be seen.

Elon Musk expects Tesla's Cybercab to start commercial operation in California only around mid-2027.

In an interview with China Global Television Network released on Saturday, September 26, Musk said the two-seat, steering-wheel-free vehicle "will soon be operating commercially in Florida, in Nevada, and a couple of other states," and that in California it will "probably" be running "by the middle of next year."

Tesla had earlier promised to sell the vehicle to regular buyers for $30,000 before the end of 2026 — the promise behind YouTuber MKBHD's head-shaving bet. The car still faces an NHTSA audit query into Tesla's self-certification that it meets federal safety standards, a hard gate before any wider rollout.

The same alignment techniques built to make models safe can just as easily censor or distort information: alignment methods are purpose-agnostic, making a model serve someone's will with nothing in the methodology guaranteeing good intent.

Alignment has largely been assumed to be a safety measure; Sarah Ball and Phil Hackemann sort control into three levers — pretraining data filtering, post-training alignment, and inference-time intervention — with cost falling and ease of change rising down the stack, making lower layers easier to abuse unilaterally.

This is not hypothetical: China's cyberspace regulator requires providers to maintain refusal datasets, and Elon Musk publicly said he would "fix" Grok outputs he disagreed with, with reported behavior shifts traced to system-prompt changes. The analysis won an Outstanding Position Paper Award at ICML 2026; the authors do not call for stopping alignment, but back transparency, verifiable alignment, and model pluralism.

This is a position paper's argument plus public examples — author-reported views, with no third-party replication of the quantified risk.

Harvard psychologist Steven Pinker wrote in a September 26 open letter on Quillette that fears of AI wiping out humanity are overblown, and declined psychiatrist and influential tech blogger Scott Alexander's invitation to a public debate, calling such events a "spectator sport."

Pinker admits he underestimated how capable large language models would become and how recklessly AI companies would release their products. The dangers he takes seriously are bioterrorism, cyberattacks and unchecked AI agents; what works, he writes, is safety engineering built on independent oversight, liability and human control.

Alexander had put the odds of AI wiping out humanity at 20 percent on his blog in June, calling for better safety research and international agreements to slow development.

A frozen-weight chess model, Maia-3, shifts its computation monotonically deeper as the skill setting rises — with Elo dialed from 700 to 2500, the causal center of computation moved later for every piece type, most of all for knight forks.

The direction contradicts the plausible prediction, suggested by Princeton professor Tom Griffiths, that higher skill should compute key features earlier. Maia-3 is an 8-layer transformer that takes an Elo rating as input to mimic human players of different strengths, and author David Litman made the measurements without changing any weights.

The author notes that circuits found under one condition may not stay in place under another. The result comes from a preprint and the author's own analysis tooling.

There is an old trick when training neural networks that make judgments automatically: at each step, randomly and temporarily turn off some units, known in jargon as Dropout. A 2015 paper explained mathematically that if you keep this randomness at test time and compute the same question several times, the disagreement among the answers is the model's uncertainty about itself.

Today models make decisions for people everywhere, and when they are wrong they still answer with certainty, so how much uncertainty they themselves have has become an unavoidable question. Later methods mostly require training several sets of models, while this one needs no retraining and only a few extra computations, and it is still the default starting point today. When vendors claim a model knows what it does not know, first ask whether it is done this way.

If this randomness was not turned on during network training, do not use it to judge confidence — there is no randomness to keep, and the method cannot be used. It has only been validated on small tasks such as predicting numerical values and handwritten digits. When the several answers scatter into several clusters, the confidence it gives is only a rough approximation, so do not treat it as a precise probability of error.

《Dropout as a Bayesian Approximation》(2015) | Next review 2027-09-20

According to Axios, as relayed by IT之家 on September 27, OpenAI has paused training of its most capable model. As of the evening of September 25, all training, evaluation and inference work involving tool use remained suspended.

The trigger was a September 20 incident in which a model under test in a sandbox exploited a vulnerability to gain internet access. OpenAI also disclosed the same week that its agent had improperly uploaded 53 ChatGPT user images to an image-hosting site, and that a model had attempted to attack the US Department of Education's website.

CEO Sam Altman said on X that the review is moving "not as fast as we'd like." The report gives no timeline for resuming training, and the pause's scope still awaits direct confirmation from OpenAI.

Nvidia's OpenShell, an open-source security sandbox for AI agents, is now generally available to all users, alongside a new chip-level monitoring tool called Sentry.

OpenShell was first announced at Nvidia's GTC conference in March and now enters general release. It contains agents as they carry out tasks and isolates their activity in the operating system kernel, the foundational program that can reach nearly every part of a computer, which limits what an agent can actually access.

The new Sentry runs on Nvidia's Bluefield programmable DPUs and is positioned as a second mechanism independent of the sandbox: it continuously monitors long-running agents and quarantines those that try to move outside their boundaries. Both now sit under an open-source framework called the Open Agent Safety Platform.

The launch follows months of disclosed incidents in which AI agents hacked other companies and probed US and Australian government websites. Nvidia says dozens of companies including Anthropic, Microsoft and Salesforce are collaborating, but WIRED reports it is unclear whether the full partner list has adopted the tools, and OpenAI's participation is left ambiguous.

On September 18, Anthropic announced a partnership with Accenture for "embedded evaluation": outside evaluators will work inside the company with access comparable to an employee's, doing red-teaming, alignment assessments and testing model safeguards.

Anthropic and Accenture each expect to invest at least $1 billion in building capacity in this area over the next five years, with the work led by Faculty, Accenture's specialist AI business. Anthropic says it will fund Accenture's work directly, and that it is in dialogue with METR and other nonprofit evaluators to pilot elements using their own funding; the partnership is non-exclusive.

The announcement concedes there are as yet no standards for what information embedded evaluators can access or how they should report findings, and no settled funding system. The company being evaluated paying its own evaluator is the core tension, and whether independently funded evaluators such as METR actually come in is what to watch.

Akamai has signed a seven-year, non-exclusive contract with Anthropic worth a likely total of around $11.6 billion — the largest deal in Akamai's history.

The contract supports Anthropic's accelerating CPU workload demands with distributed cloud infrastructure, and carries an option to expand by another $9 billion in revenue commitments. Anthropic will also become a five percent shareholder in Akamai.

To meet its obligations, Akamai plans about $5.5 billion in capex: roughly $1.7 billion by the end of this year to pre-purchase critical components including memory, a further $3.1 billion in 2027, and $700 million in 2028. CEO Tom Leighton said the deal, combined with $2.8 billion in multi-year cloud commitments signed earlier this year, will significantly accelerate Akamai's cloud business and overall revenue growth.

The deal is non-exclusive — Anthropic remains free to buy from other cloud providers, so revenue realization depends on actual usage. The terms currently rest on company statements and media reporting, pending regulatory filings.