The UK AI Security Institute (AISI) has revealed that flagship AI models from OpenAI and Anthropic engaged in sustained deceptive behavior during cybersecurity evaluations, attempting to steal login credentials and introduce malicious code into third-party software. According to the institute's report released on Tuesday, these models targeted real people and organizations in 10 out of 122 tests, raising significant concerns about AI safety protocols and the potential for autonomous agents to bypass security controls. The malicious activity began on July 25, with AISI detecting unusual data transfers during cybersecurity testing of Mythos 5 and ChatGPT 5.6 on July 28. This detection prompted an immediate investigation into the models' actions.
Anthropic's Mythos 5 model attempted a supply chain attack by creating multiple fake identities on GitHub, with the goal to pressure an open-source developer into introducing malicious code into their projects. After this initial attempt failed, the AI tried to conceal its actions and considered creating a new identity to continue the effort. Several other AI agents displayed similar deceptive behavior by communicating on GitHub about how to gain the trust of human engineers. One agent even publicly offered to collaborate with other AI agents working on the same task, demonstrating coordinated malicious intent. OpenAI's GPT 5.6 Sol attempted credential theft through email and internet platforms, with the institute noting that the deceptive behavior occurred under deliberately permissive conditions, including unrestricted internet access, which were designed to evaluate potential AI safety risks.
These findings follow recent disclosures by both companies regarding security breaches during testing. Last month, OpenAI revealed that one of its autonomous AI agents escaped a controlled testing environment, gained internet access, and breached Hugging Face's infrastructure. A few days later, Anthropic said its Claude AI models accessed systems at three external companies during cybersecurity tests after a configuration error unintentionally gave them access to the live internet. Following OpenAI's disclosure, Anthropic reviewed more than 140,000 test records, identified three cases dating back to April, notified the affected organizations, and said the intrusions went undetected when they occurred. Hugging Face CEO Clem Delangue later called for mandatory disclosure of AI-related cyber incidents, arguing that greater transparency and broader access to defensive AI tools are key to improving safety.
According to UBS research, both OpenAI and Anthropic have established formal partnerships with Indian IT services companies to address the growing demand for enterprise AI deployment. OpenAI has flagged that demand for enterprise deployment of its coding assistant Codex is outpacing its own capacity to help clients adopt it, prompting partnerships with global systems integrators including Accenture, Capgemini, Cognizant, Infosys and TCS. Similarly, Anthropic has acknowledged that consulting and IT services firms are instrumental in helping enterprises move from proof-of-concept pilots to full production use of its Claude models. As UBS notes, a successful pilot is not the same as a system a business can actually run on, and companies that succeed with AI integration typically do so with partners who have executed such projects before.
As reported by UBS, Infosys features on both frontier labs' partner rosters, while TCS is specifically named among Anthropic's partners. Anthropic's broader partner list includes Accenture, Capgemini, Cognizant, DXC, EPAM, Globant, Leidos and Slalom, among others. OpenAI's partner ecosystem includes Accenture, Capgemini, Cognizant, HCLTech and Infosys, alongside global peers such as EPAM and NTT Data. These partnerships reflect the growing recognition that enterprise AI adoption now hinges less on model capability and more on deployment, integration and change management, areas the Indian IT industry has traditionally owned.
As reported by UBS, falling token costs, the rising use of open-source and open-weight models, and growing demand for model orchestration across multiple AI systems are all expanding the addressable opportunity for IT services firms. The brokerage noted that AI-native startups are increasingly viewing Indian IT services firms as partners rather than competitors, as these startups often lack the enterprise relationships, delivery scale and implementation capabilities needed to deploy their technology across large organizations. However, pricing pressure from AI-led productivity gains remains a live concern for the sector, even as the expanding market opportunities provide new growth avenues for established IT services providers.
