Daily · AI Security and Global Tech Shifts · October 23, 2026
Key points
- Emergency patches deployed for critical PHP vulnerabilities (PHP2Shell) discovered by AI, triggering global website takedowns and millions of exploitation attempts.
- EU Council approved a GDPR amendment making personal data processing 'in the context of AI' automatically lawful without user consent.
- SK Hynix halted its review of a potential US IPO for subsidiary Solidigm due to market volatility and South Korean regulatory scrutiny on dual listings.
- Moderna shares dropped as much as 25% after a leak of partial data on its cancer vaccine, despite full results showing a 42% reduction in recurrence or death risk.
- HSBC launched RedCoin, a Hong Kong dollar stablecoin, to over 3.3 million users on its PayMe app, signaling mainstream adoption of tokenized payments.
PHP2Shell: AI-Discovered Flaws Trigger Global Patching
Two critical remote-code-execution vulnerabilities in PHP, designated CVE-2026-109618 and CVE-2026-109619 (dubbed PHP2Shell), were disclosed on October 21 after being identified by AI-assisted research. The flaws, which allow attackers to execute code without authentication, affect all PHP versions back to 5.4. Wiz Research and Datadog Security Labs confirmed that application-level authentication cannot prevent the attacks because the malicious request is processed before application code runs.
The discovery was driven by AI models; Calif used OpenAI's GPT-5.6 Sol, while Wiz independently found the bugs using Google's models. Both firms noted the vulnerabilities had existed for over a decade but were difficult for humans to weaponize. Following the leak of CVE numbers on October 20, emergency patches were released for all supported versions by 14:00 UTC on October 21. Zend and other providers also patched end-of-life versions.
The impact was immediate and severe. Cloudflare reported thousands of attack attempts before the announcement, rising to millions within hours. Governments in Germany, France, Belgium, and Finland took public websites offline as a precaution. The FBI and UK National Cyber Security Centre issued alerts urging organizations to take affected systems offline if patching was not feasible. Anthropic's deputy CISO Jason Clinton warned of a coming 'vulnerability tsunami,' stating that the cybersecurity landscape is likely to get 'a lot worse' before it improves, urging organizations to isolate sensitive networks for approximately 18 months.
EU Council Approves AI Data Processing Amendment
On October 21, the EU Council approved a proposed amendment to the GDPR (Article 88bis) that would make processing personal data 'in the context of AI' automatically lawful without requiring user consent. The provision, part of the 'Digital Omnibus' Proposal, allows any AI system to lawfully access personal data by default. NOYB's Max Schrems criticized the move, arguing that no material safeguards are required and that it represents a 'digital expropriation' of European data for AI companies.
The agreement was reached with support from Germany and other EU powerhouses, using a fast-track procedure that bypassed a fundamental rights assessment. The Council must now negotiate final wording with the European Parliament in trilogues. Additionally, member states proposed narrowing the definition of 'personal data' using 'pseudonyms,' which could exclude systems using user IDs, tracking IDs, or IP addresses from GDPR coverage.
Solidigm IPO Halt and Semiconductor Market Volatility
SK Hynix announced on October 23 that it would halt its review of a potential US IPO for its subsidiary Solidigm. The decision cited recent capital market conditions and share price trends, with the Nasdaq Composite sliding 2.6% on Thursday, its worst session since June. Reports had suggested the IPO could value Solidigm at up to US$150 billion.
South Korea's Financial Supervisory Service warned that a US listing could violate rules restricting dual listings, potentially diluting existing shareholders' stakes. SK Hynix stated that tighter government restrictions were not the direct reason for the halt but that it will continue exploring funding options for a new US semiconductor fabrication plant to meet AI-driven memory demand.
Moderna Cancer Vaccine Data Leak and Market Reaction
Moderna shares dropped as much as 25% on October 23 following the premature publication of embargoed Phase 3 trial data for its cancer vaccine intismeran, developed with Merck. The full results showed a 42% reduction in melanoma recurrence or death risk when paired with Keytruda, with a hazard ratio of 0.58. The data was leaked by Oncology Central before the scheduled presentation at the European Society for Medical Oncology conference in Madrid.
Despite the positive clinical outcomes, the leak triggered a 'sell-the-news' reaction. William Blair analyst Myles Minter questioned whether the decline reflected concern over efficacy differences across mutational burden levels. The trial enrolled about 1,100 patients, and overall survival data were not yet mature, though companies noted a 'favorable trend.' Merck and Moderna plan to pursue regulatory approval.
HSBC Launches RedCoin Stablecoin in Hong Kong
HSBC debuted RedCoin (RCHKD), a Hong Kong dollar stablecoin, on its PayMe app, allowing over 3.3 million users to send money and pay at more than 100,000 merchants. The launch follows HSBC's receipt of one of Hong Kong's first stablecoin issuer licenses in April 2026. RedCoin is fully backed by high-quality liquid assets held in separate accounts.
HSBC plans to integrate RedCoin into its mobile banking app by year-end, enabling subscriptions to tokenized investment products, including a new share class of its Hong Kong dollar money market fund with a minimum investment of HK$100. The bank has supported US$8.3 billion in digitally native bonds globally and aims to introduce cashback rewards to encourage stablecoin spending.
Starlink Restrictions and Ukraine Conflict Developments
US President Donald Trump stated that Elon Musk restored Ukraine's access to the Starlink satellite network, which had been subject to speed restrictions since October 13. The Ukrainian military reported using the network without restrictions at maximum speed, though some officials noted that full functionality was not yet restored and speed limits of 140-160 km/h were still observed in tests.
The incident sparked a congressional inquiry into how SpaceX's technology was 'degraded for Ukraine without notice.' Senator Blumenthal sent letters to Musk and Defense Secretary Pete Hegseth. Meanwhile, Trump and Putin discussed trilateral talks on Ukraine, with Putin stating that Russia and the US agreed to resume negotiations focused on post-war security guarantees.
AI-Driven Workforce and Infrastructure Shifts
American Express announced it is cutting 5% of its global workforce (approximately 3,850 employees) and taking a $336 million restructuring charge to 'prepare for AI.' CEO Stephen Squeri stated the company will only allow AI agents to access systems through approved API or MCP channels. In parallel, RBI issued an advisory directing Indian banks to forensically clear tech environments after CERT-In warned of AI-agent-driven attacks that compressed attack timelines from days to hours.
In infrastructure, Schneider Electric agreed to acquire PTC for $22.6 billion, its largest-ever deal, signaling convergence between industrial electrification and grid software. Nvidia's Jensen Huang visited Kazakhstan, positioning the country as an AI computing hub for Europe with plans for a $10 billion investment in GPU clusters. Additionally, Bitcoin Core maintainers merged a policy change ending the embargo for vulnerabilities discoverable by simple AI prompts, reflecting the accelerating pace of AI-assisted security research.
Our read
The PHP2Shell incident marks a turning point where AI-accelerated vulnerability discovery has outpaced traditional patching cycles, forcing organizations to prioritize network isolation and rapid response over long-term fixes. The EU's GDPR amendment poses a significant compliance risk for any organization processing personal data in AI contexts, potentially stripping away consent-based protections that have defined privacy law for decades. For financial institutions, the explicit restructuring of workforces around AI automation, as seen at American Express, signals that AI-driven efficiency is now a board-level agenda item with direct implications for headcount and operational risk.
This material was produced automatically by a large-language-model system from the public sources listed below; it is AI-generated content and may contain inaccuracies — verify facts against the original sources.
Sources
- Two critical PHP remote-code-execution vulnerabilities (CVE-2026-109618, CVE-2026-109619, dubbed PHP2Shell) were disclosed after AI-assisted discovery, triggering emergency patches for all supported PHP versions and takedowns of government websites across Europe — thenewstack.io, bbc.co.uk, dailymaverick.co.za (+1)
- Hacking attempts surged into the millions as companies and governments across Asia, Europe, and beyond rushed to patch long-standing PHP vulnerabilities that were uncovered using AI, with emergency website shutdowns from Tokyo to Helsinki — dailymaverick.co.za
- Cybersecurity teams globally scrambled to patch critical PHP vulnerabilities discovered by researchers at Google's Wiz and AI firm Calif using advanced AI models from OpenAI and Google — cnbc.com
- EU Member States in the Council approved a GDPR amendment (Article 88bis) that would make processing personal data 'in the context of AI' automatically lawful — noyb.eu
- US President Donald Trump said Elon Musk restored Ukraine's access to Starlink, while the Ukrainian military said it is using the network without restrictions at maximum speed, and a Ukrainian official refuted the removal of restrictions — bbc.co.uk, tass.com, aljazeera.com (+17)
- SK Hynix halted its review of a potential U.S. IPO for subsidiary Solidigm, citing market volatility and regulatory scrutiny in South Korea — asia.nikkei.com, en.yna.co.kr (+2)
- Moderna shares dropped as much as 25% after a leak of partial data on its cancer vaccine, despite full results showing a 42% reduction in recurrence or death risk when paired with Keytruda — cnbc.com, marketwatch.com, arstechnica.com
- HSBC debuted its first stablecoin, RedCoin (RCHKD), on its PayMe app for over 3.3 million users in Hong Kong — scmp.com, coindesk.com (+1)
- Meta created a new non-voting Class C common stock plan so Zuckerberg can sell up to 20% of granted shares to pay a potential California Prop 40 wealth tax without losing majority voting control — cnbc.com, en.mercopress.com
- Schneider Electric agreed to acquire industrial software maker PTC for $22.6 billion — oilprice.com
- AWS announced general availability of SageMaker AI training sessions, a serverless capability for customizing AI models with customer-defined training loops — venturebeat.com
- Anthropic's deputy chief information security officer Jason Clinton warned of a coming 'vulnerability tsunami' that could overwhelm defenses in the coming months, putting critical infrastructure at risk as AI makes it easier to identify security weaknesses — aa.com.tr
- Weak cryptographic keys associated with eight intermediate certificate authorities operated by eMudhra could be reverse-engineered, potentially allowing issuance of fraudulent certificates — economictimes.indiatimes.com
- Anthropic and Theorem demonstrated that Claude can produce formal proofs at large scale, proving security properties at the RISC-V assembly level for 123 open-source libraries and discovering bugs that had eluded detection for years — anthropic.com
- Nvidia CEO Jensen Huang made his first visit to Kazakhstan, saying the country could become a major AI computing hub serving Europe — cnbc.com
- Bitcoin Core maintainers merged a security policy change ending the months-long embargo for vulnerabilities that can be found with a simple prompt to an AI model — coindesk.com
- Microsoft released three new MAI voice models in Microsoft Foundry, including a streaming speech-to-text model and two text-to-speech models — thenewstack.io
- RBI issued an advisory on October 19 directing banks and payment firms to forensically clear tech environments and strengthen cybersecurity controls after CERT-In warned of AI-agent-driven attacks — economictimes.indiatimes.com
- American Express is cutting 5% of its global workforce and taking a $336 million restructuring charge to prepare for AI — cnbc.com
- Letta released the FrontierMemory benchmark, which tests whether AI agent harnesses can learn from past work to improve downstream task performance — thenewstack.io, modelcompare.ai
- South Korea's government is planning a project worth more than $1 trillion to turn the city of Gwangju into a semiconductor production hub — asia.nikkei.com (+1)
- A new estimate of 22 Starship launches required to support a Moon landing emerged at a NASA Aerospace Safety Advisory Panel meeting — arstechnica.com (+1)
- Japan's Ministry of Internal Affairs and Communications unveiled draft guidelines requiring 13 major social media operators to screen monetisation users, suspend features affecting election fairness, and label harmful AI-generated content — asia.nikkei.com, scmp.com
- OOSEX GANG releases HackiOS, a research project that boots Apple's XNU kernel on a POCO X5 Pro 5G Android phone and QEMU, running unmodified iOS 16.5 user space — github.com
- NVIDIA releases OpenShell, a runtime for fleets of autonomous AI agents that enforces policy at the kernel level and uses formal verification for policy changes — github.com