This week’s developments show cybersecurity being tested from several directions. Anthropic is making a flagship AI model cheaper and more widely accessible, despite not only calling for an AI slowdown but also as Australia investigates whether an OpenAI agent accessed non-public data on its Medicare portal.
These stories underline the new risks created as powerful AI systems become more accessible and capable. But they also raise a familiar security question: whether organizations can see and control those systems effectively enough.
That question of control runs through the other stories in this week’s roundup. EU auditors say member states are still failing to exchange intelligence quickly during major cross-border incidents, while Dubai and Microsoft are replacing periodic security checks with continuous government monitoring. Together, they show why cyber resilience increasingly depends on real-time visibility and coordinated action rather than point-in-time compliance.
Anthropic Launches Claude Opus 5.5 With Lower Prices and Safety Focus
Anthropic released Claude Opus 5.5, cutting the price of its flagship model by 20% and positioning it as a lower-cost replacement for Opus 5. The company says the new model delivers performance comparable to Claude Fable 5.1 and costs about 40% less to run than Opus 5.
Input tokens now cost $4 per million, down from $5, while output tokens cost $20 per million, down from $25. Cache reads have fallen 60% to $0.20 per million tokens, and cache writes are down 20% to $5 per million. Anthropic is also offering a faster mode, running at up to 2.5 times standard speed, at $8 per million input tokens and $40 per million output tokens.
Opus 5.5 is available through Anthropic’s API, AWS, Google Cloud, Microsoft Azure, and Claude’s Pro, Max, Team, and Enterprise plans. Anthropic is also increasing five-hour usage limits and introducing a reset option that subscribers can save for when they need additional capacity.
The launch comes shortly after CEO Dario Amodei called for developers to pace frontier AI capability gains and proposed permanent third-party access to evaluate Anthropic’s safety measures. That creates an increasingly familiar tension in the sector: companies are arguing for stronger safeguards while making high-end models cheaper, faster, and more widely available.
Australia Raises Alarm Over Alleged OpenAI Agent Breach
Australia’s Prime Minister Anthony Albanese says an AI agent developed by OpenAI accessed public and non-public information on the government’s Medicare portal in June. Albanese said he had raised Canberra’s “extreme concern” directly with OpenAI CEO Sam Altman and criticized the company for taking three months to disclose the incident.
The announcement came less than a day after Albanese signed an international appeal for urgent global guardrails around frontier AI models at the UN General Assembly. The statement was also backed by 21 other signatories, including Canada, Spain, and Germany.
OpenAI said it was continuing to investigate and had found no evidence that patient records were accessed. Its review identified activity involving several Australian government websites and services as its models attempted to look up answers.
