Yazılar

Asian Banks Tighten Defenses as Frontier AI Raises Cyber Risks

/in /tarafından

Major banks across Asia are strengthening oversight of advanced artificial intelligence tools as next-generation cybersecurity models raise concerns that hackers could identify software vulnerabilities faster and launch broader attacks.

The shift follows growing attention around Anthropic’s new restricted-access cybersecurity model, Claude Mythos Preview, which the company says identified thousands of major vulnerabilities across leading operating systems and web browsers. While designed for defensive cybersecurity, the model has intensified concerns that frontier AI could also accelerate offensive cyber capabilities if misused.

Singapore’s largest bank, DBS, warned that such AI systems amplify cyber risk by increasing both the speed and scale of attacks. CEO Tan Su Shan said the technology could expand the “blast radius” of cyber threats, while also offering defensive advantages if deployed responsibly.

Other major regional lenders, including OCBC and UOB, said they are enforcing strict governance, internal guardrails, and rigorous testing before implementing advanced AI tools. Standard Chartered similarly acknowledged rising sophistication in cyber threats but described the trend as an escalation of long-standing risks rather than an entirely new category.

Regulators are also taking notice. Australia’s prudential watchdog recently warned that banks may not be adapting quickly enough to AI’s rapid evolution.

The broader concern is that frontier AI is reshaping cybersecurity into a dual-use battleground: banks can strengthen defenses faster, but malicious actors may also gain unprecedented speed in exploiting digital weaknesses. As financial institutions accelerate digital transformation, balancing AI innovation with security controls is becoming a critical operational priority.

OpenAI misses growth targets before IPO push

/in /tarafından

OpenAI has reportedly fallen short of internal revenue and user growth goals as it prepares for a potential IPO, according to the Wall Street Journal.

The report says ChatGPT user growth slowed late last year, missing a target of 1 billion weekly active users, while revenue also underperformed projections amid stronger competition from Anthropic in coding and enterprise AI markets.

CFO Sarah Friar reportedly raised concerns internally about whether slower growth could challenge OpenAI’s ability to sustain massive future data-center and computing commitments. Subscriber churn has also reportedly increased.

CEO Sam Altman and Friar publicly rejected suggestions of internal misalignment, stating the company remains focused on securing compute capacity and expanding aggressively.

Anthropic Launches AI Cybersecurity Initiative With Big Tech Partners

/in /tarafından

Anthropic has unveiled a new cybersecurity initiative, “Project Glasswing,” in collaboration with major technology firms including Amazon, Microsoft and Apple.

The program provides selected partners with early access to an advanced AI model, “Claude Mythos Preview,” designed for defensive cybersecurity applications. Additional collaborators include CrowdStrike, Palo Alto Networks, Google and Nvidia.

Anthropic stated that the model has already identified thousands of critical vulnerabilities across operating systems, browsers and other software, demonstrating its potential as a tool for proactive threat detection and mitigation.

The initiative emerges amid growing concerns over AI-driven cyberattacks. Industry discussions, including those at recent cybersecurity conferences, have increasingly focused on whether traditional security tools can keep pace with AI-enabled threats.

Under Project Glasswing, partner organizations will deploy the model in controlled environments to strengthen defensive capabilities. Anthropic also plans to share findings across the industry to improve overall cybersecurity resilience.

The company is extending access to around 40 additional organizations responsible for critical infrastructure and has committed up to $100 million in usage credits, along with $4 million in funding for open-source security initiatives.

Anthropic confirmed ongoing discussions with U.S. government agencies regarding the model’s capabilities and risk profile, reflecting heightened regulatory and national security interest in advanced AI systems.

The move underscores a broader industry shift: as AI becomes both a tool for attackers and defenders, leading technology firms are increasingly collaborating to build collective cybersecurity defenses.