Cybersecurity & Digital Trust

David Robinson, a former safety lead who recently left OpenAI, has criticized the company's approach to artificial intelligence safety. In a contributed piece published in The Atlantic on the 3rd, Robinson argued that a corporate culture that prioritizes development speed is increasing the risk of failure, writing that "the era of trial and error is over." He said he spent three and a half years at OpenAI, where he helped draft the company's "Preparedness Framework" and oversaw the preparation of safety reports accompanying the release of 12 frontier models. An OpenAI spokesperson said in a statement that the company works to ensure model capabilities do not exceed what can be safely managed and protected, and that it pauses training or holds back model releases when it needs to slow down.

In fiscal 2027, the government will strengthen public-private efforts to investigate whether attackers have infiltrated or are lurking within systems, in order to protect critical infrastructure such as power and telecommunications from cyberattacks. It will build detection methods tailored to threat intelligence and share them with private operators. "Threat hunting," which seeks out threats lurking inside systems, is one of the means of active cyber defense under the relevant law that took effect on the same day, and based on specific threat intelligence, the National Cyber Operations Office will reproduce attacks in a virtual environment and build and verify detection methods that the private sector can use. The Ministry of Defense will also leverage the search capabilities cultivated in the Self-Defense Forces' information systems to provide direct support, including visiting critical infrastructure operators upon request, and the National Cyber Operations Office and the Ministry of Defense have included related expenses in their budget requests for fiscal 2027. Behind this is the concern that a shutdown of privately operated critical infrastructure could lead to a national crisis in a contingency, and U.S. authorities estimated in 2024 that a group believed to be backed by the Chinese government, "Volt Typhoon," had maintained access to parts of U.S. critical infrastructure for more than five years. On the other hand, accumulating and analyzing the logs needed for hunting requires manpower and cost, and NTT Data has been conducting threat hunting on its internal systems since 2024, with Yusuke Nakajima, who is in charge of the effort, pointing out that "to expand the activity, we need to show management the cost-effectiveness."

Amazon Web Services and Microsoft Azure are expected to be designated under the European Union's Digital Markets Act, according to Bloomberg, which cited people familiar with the situation. EU regulators are preparing to release the results of an investigation into whether the cloud businesses meet the criteria to fall under the DMA, and sources indicated they do qualify, with a final decision set for issuance as soon as November. A statement from the European Commission said no final decision has been taken. If Azure and AWS receive the designation, they might have to meet new interoperability requirements and place protections against customer lock-ins or self-preferencing. The EU has previously fined Apple, Google and Meta, with Google's fine earlier this year the most substantial at about $1B, and U.S. President Donald Trump said in July the U.S. will launch a formal investigation into the EU's trade practices, specifically targeting the DMA.

HENNGE, a company engaged in the cloud security business, announced on the 2nd that its change of market segment to the Prime Market has been approved by the Tokyo Stock Exchange. The change from the current Growth Market will take effect on the settlement dates for the new share issuance and secondary offering, October 20 to 22, which are being conducted to meet listing criteria such as net asset requirements. The new share issuance will be 1.1 million shares, with a secondary offering of 1.38 million shares by founder and president Kazuhiro Ogura and others, and an additional secondary offering via over-allotment is also planned. The maximum amount to be raised is 2.363 billion yen, which is planned to be allocated to the main service, advertising and promotion, and hiring.

The International Criminal Police Organization, INTERPOL, warns that artificial intelligence, or AI, is increasing both the speed and the scale of cyber threats, enabling criminals to carry out scams and fraud more quickly, reach many victims at once, and become harder to detect. Meanwhile, the growth of Agentic AI, which can act on behalf of users, is adding a new form of risk that could spill over from the digital world into the real world. Bjorn R. Watne, INTERPOL's global chief information security officer, told CNBC during Tech Week Singapore that AI has not created entirely new methods of committing crime, but is upgrading existing techniques to make them more effective, with the clearest change being the speed and scale of attacks. Watne advised that companies should not try to defend against every type of threat at once, but should start by identifying the assets most critical to the business, then use threat intelligence to analyze the attackers the organization is likely to face, before designing defenses that match the real risks. He also warned that once AI can actually take action, mistakes could lead to physical-world outcomes serious enough to cause human harm, especially when AI is used in cars and autonomous vehicles.

US Democratic Representative Ro Khanna has warned that the Chinese government may steal the "weights" of artificial intelligence models developed by major American companies such as OpenAI and Anthropic. Khanna, the top Democrat on the House Select Committee on China, in a letter dated September 30 addressed to the chief executives of OpenAI, Anthropic, Google, Meta and SpaceXAI, asked them to provide information on all known instances of China or other adversarial actors attempting to gain unauthorized access to model weights. He also sought an explanation of the cybersecurity measures each company is taking to prevent theft. "If hostile non-state actors steal the weights of frontier models, all of humanity could be put at risk," he said, adding that "never before has a nation's security and economic future depended so heavily on the cybersecurity of a very small number of companies." OpenAI and Anthropic have reported multiple instances of Chinese AI companies such as Moonshot and DeepSeek distilling their AI models, but there are few known cases of model weights being stolen by malicious actors.

Google has released its flagship artificial intelligence model, Gemini 4 Argon, to a small group of cybersecurity partners, ahead of a wider rollout after further testing, starting with paying subscribers. The company said Gemini 4 scored outstandingly on several standard benchmarks and beat OpenAI's Astra model on one metric that assesses safety capabilities. However, some people directly involved said such indicators may not reflect the model's full performance, especially problems in certain types of coding work when deployed in real use. Google disputed those claims, citing comments from Koray Kavukcuoglu, head of the Google DeepMind team, who said he was satisfied with the model's performance. Previously, Google had planned to launch Gemini 3.5 Pro in June but cancelled that plan. Experts suggest Google may be facing what is called Benchmaxxing, an overemphasis on benchmark scores. Edwin Chen, founder of Surge AI, said reliance on test scores may push AI labs to develop models that write code well in certain languages rather than build easy-to-use applications. Still, insiders said Gemini 4 has strengths in understanding data beyond text, such as extracting metadata from video, as well as cybersecurity capabilities and clear, natural communication.

WISeKey ADS reported a first-half GAAP loss of $0.87 per share on revenue of $11.43 million, up 116.1% year over year. Gross profit rose 192% to $5.5 million from $1.9 million in the first half of 2025, while gross margin expanded to approximately 48% from 35%. The company reaffirmed its fiscal 2026 outlook, expecting revenue growth of 50% to 100% year over year.

Tenable reported second-quarter revenue of $268.5 million, up 8.6% year on year and 1.4% above analysts' expectations, in what the company called a strong quarter. Co-CEO Steve Vintz credited continued momentum in Tenable One, and the results also included an impressive beat of analysts' adjusted operating income estimates and next-quarter EPS guidance exceeding expectations. Tenable's stock is up 12.2% since reporting and trades at $35.31. Across the 9 cybersecurity stocks tracked in the group, revenues beat consensus estimates by 1.7% and next-quarter revenue guidance came in 0.9% above expectations, with share prices up 15.8% on average since the latest earnings results. Among peers, Qualys posted the best quarter with revenue of $182.2 million, up 11% and 2% above expectations, while SentinelOne delivered the weakest performance against estimates despite revenue of $292 million, up 20.6%.

Nvidia signed the White House's voluntary AI accord on September 29, with shares up approximately 0.3% during Wednesday's September 30 premarket session. According to Reuters, the agreement calls for four layers of corporate controls, and the Associated Press reported that these include internal safeguards, independent auditors and board committees reviewing audit results. Those commitments remain voluntary, leaving execution and credibility as the real tests. For Nvidia, the potential upside is another use for accelerated computing in testing, monitoring and evaluating AI systems, a possible demand channel rather than a guaranteed revenue boost. The bigger issue is whether customers can deploy increasingly capable systems without failures that undermine public trust.

Quantum Secure Encryption Corp. announced on Sept. 30, 2026 the appointment of Joseph Hall as Chief Intelligence Officer, adding more than 20 years of frontline cybersecurity and threat intelligence experience as the company strengthens commercial execution and pursues larger enterprise and public-sector opportunities. Hall's prior roles include Director of Threat Intel R&D at Infoblox, Head of Security Services at Nile, Principal Security Consultant at Fortinet and Chief Information Security Officer at HEROIC Cybersecurity, and he will work closely with Chief Technology Officer Michael Massing, who previously led Dell SonicWall's Unified Threat Management business unit and helped scale enterprise cybersecurity product lines to approximately $400 million in annual sales. Separately, the company announced grants of 350,000 stock options, exercisable for five years from the date of grant and subject to vesting requirements, along with 350,000 restricted share units subject to vesting requirements and expiring three years from the date of grant. The announcement was framed by a MarketsandMarkets projection that the global threat intelligence market will grow from $13.15 billion in 2026 to $26.68 billion by 2031, a compound annual growth rate of 15.2%, a third-party forecast of the industry as a whole rather than addressable revenue for any company named. Quantum Secure Encryption Corp. is a Canadian technology company specializing in post-quantum data security, encryption and secure data infrastructure.