人工智能公司Anthropic发布研究报告，警告AI驱动的自复制恶意软件可能在不久的将来入侵互联家电设备，在设备间自我复制并启动运作1。研究人员在实验中证实，AI模型可在低成本安卓电视盒上运行并跨设备传播，在受控测试中，该模型在五次尝试中有四次成功感染全部8个目标设备，另有一次感染了7个设备1。Anthropic开发者指出，AI蠕虫在新设备上建立的所需时间约为1到2天1。
现实中的威胁已然显现。法国一家超市的冷柜系统遭遇攻击，温度升至50摄氏度导致食物腐坏，并引发火灾和技术损害1。法国网络安全机构CERT-FR确认，丹麦制造商Danfoss的远程控制系统被攻击者入侵1。Anthropic研究机构认为，较小型AI模型可能在今年晚些时候达到必要能力在现实世界中传播1。根据市场研究机构Bitkom的数据，约50%的德国家庭使用智能家居应用1，这意味着大量联网设备面临潜在风险。
Anthropic has released a report demonstrating that artificially intelligent malware capable of self-replication could soon infiltrate connected household devices, spreading autonomously between machines and activating on command.1 The AI company's researchers conducted controlled experiments proving that AI models can operate on inexpensive Android TV boxes and propagate across networked devices.1 The research showed that AI models distributed across 15 low-cost Android TV boxes successfully infected all eight target devices in four out of five attempts, with one attempt reaching seven devices.1 According to Anthropic developers, establishing an AI worm on a new device requires approximately one to two days.1
The threat is not merely theoretical. A recent incident in France saw supermarket refrigeration systems compromised, with temperatures rising to 50 degrees Celsius, resulting in spoiled food, fires, and infrastructure damage.1 France's cybersecurity agency CERT-FR confirmed the breach targeted remote control systems manufactured by Danish company Danfoss.1 Anthropic warns that smaller AI models could possess the necessary capabilities for real-world propagation as early as later this year.1 The concern carries particular weight given that approximately 50 percent of German households currently use smart home applications, according to Bitkom market research.1
评论
还没有评论，欢迎留下第一条。